Voice: Voicemail Handling and Québec French
Your voice agents now know what to do when nobody picks up. Voicemail detection is on for every voice agent, and the default behavior is to hang up rather than keep talking into an answering machine. If you would rather leave a message, you do not need to configure anything new: describe what your agent should say in a voicemail anywhere in its instructions or in one of its Procedures, and your agent uses it, in whichever language the call is running in.French voice agents also sound noticeably more local. Agents speaking French now use a Québécois accent rather than a Parisian one, including how names, numbers, and regional terms are pronounced, which makes calls to Quebec contacts feel considerably less foreign.
- Voicemail detection on every voice agent, with hanging up as the safe default
- Voicemail messages taken from your existing instructions or Procedures, with nothing new to configure
- Language-aware voicemail so an English call leaves an English message even on a French-first agent
- Québec French for voice agents, covering accent and pronunciation
- Cleaner call transcripts in your Inbox
- Infinite scroll on the sidebar Inbox and Contacts lists
Features
- Voicemail detection is enabled automatically for all voice agents, inbound and outbound
- Describe your voicemail message in your agent’s instructions or in a Procedure and your agent leaves it when it reaches an answering machine
- Voicemail messages are available in every language your agent publishes, and the right one is chosen per call
- French voice agents use a Québécois accent and pronunciation rather than defaulting to Parisian French
- The sidebar Inbox and Contacts lists load continuously as you scroll instead of stopping at a fixed number of conversations
Bug Fixes
- Outbound calls now use the first message and the language configured for the batch
- Voice control tags no longer appear in call transcripts in the Inbox
- The agent publish button no longer stays disabled after a slow publish
Procedures: Playbooks, Contact Timelines, and Enrichment
Not every rule your agent needs belongs in its instructions. Discount requests, refund rules, warranty questions, after-hours calls, escalation paths, compliance scripts: each one matters, but each one only applies some of the time, and packing them all into one set of instructions dilutes every conversation. Procedures solve this. You describe a situation once and the steps your agent should follow when it comes up, and your agent pulls up that playbook only when a conversation actually matches it.Your agent keeps a short index of the procedures you have written and loads the full playbook on demand, so you can write long, detailed instructions for a long tail of scenarios without slowing anything down. The mechanism stays invisible to the person your agent is talking to, and you can see exactly which procedure was used in the Events timeline of any conversation.Contacts also became a great deal more transparent. Every contact now has an activity timeline showing what happened to them and when, and attribute values show where they came from. Workflows can react to contacts directly through new Contact Created and Contact Updated triggers, and a new Enrichment action fills in missing contact attributes from external data.
- Procedures, scenario-specific playbooks your agent loads only when the situation calls for them
- Contact activity timeline covering insights collected, attribute changes, and every interaction
- Contact triggers for workflows, so a new or updated contact can start a workflow
- Enrichment action to fill in missing contact attributes from external data
- Workflow run costs visible on the runs table, and node messages that can be written as a prompt
- Completions endpoint for the Apps API, for one-shot generation with no conversation
- New Input design for the Messenger widget
Features
- Your agent loads a procedure only when a conversation matches it, and never reveals procedure names or contents to visitors
- The Events timeline of any conversation shows which procedure was loaded
- Procedures work on every channel, including chat, forms, email, SMS, and voice calls
- Procedures follow the same draft and publish lifecycle as the rest of your agent’s configuration
- Ask Dimedove can read your existing procedures and draft new ones for your approval
- Contact activity timelines cover contact creation, blocking, insights collected, attribute updates with their previous value and source, and each interaction per channel
- Existing contacts have a populated activity timeline from day one
- Attribute values now show their source, including values filled in by enrichment
- New Contact Created and Contact Updated workflow triggers, plus a Facebook Message Received trigger, with contact attributes available to downstream nodes
- The Enrichment action lets you choose which attributes to enrich, add optional guidance per field, and decide whether to overwrite existing values, and runs in the background so your workflow does not wait on it
- Workflow node message bodies can be written as a prompt and composed at run time from the trigger data, instead of written out literally
- A new completions endpoint at
POST /apps/{app_id}/completionsgenerates a single piece of text with no conversation and no history, for autocompleting a field, drafting a message, or summarizing content in your agent’s voice, with support for streaming, knowledge base grounding, a low-latency mode, and metadata - The Messenger widget gained an Input design alongside Compact, Wide, and Full screen, rendering as a slim composer that expands into the chat panel, with a placeholder that cycles your agent’s conversation starters
- Credit alerts let you set your own thresholds and receive an email when your team’s credit usage crosses one
Notes
- Procedures require a paid plan
- The Enrichment action consumes credits based on the data retrieved
Ask Dimedove: Your Workspace, in Conversation
Everything your agents do lands somewhere in Dimedove: conversations across every channel, contacts and the insights collected about them, agent configurations, workflow runs. Ask Dimedove lets you interrogate all of it by asking. Search and read across conversations, look up a contact and what your agents learned about them, count and group by channel or period or status, check how an agent is configured, and get the answer back as a chart when a chart is what you need.It also changes things for you, but never on its own. When you ask Ask Dimedove to rewrite an agent’s instructions, adjust its tone, or update how it moderates comments, it comes back with a before and after comparison that you approve or reject. Approved changes to an agent land as a draft and go live through the normal versions flow, so nothing reaches your customers without you publishing it. If your request is ambiguous across several agents, it asks which one you meant rather than guessing.You can also take your workspace with you. A new Ask Dimedove connector exposes your workspace to an external MCP client such as Claude Desktop or Cursor, so you can query your Dimedove data from inside the tools you already work in. The connector is read-only by design: the tools that propose or write changes are not exposed through it.
- Ask questions about your workspace across conversations, contacts, agents, and channels
- Charts on request for counts, groupings, and trends
- Configuration changes proposed, never applied, with a before and after comparison you approve
- Approved agent changes land as a draft, published through the normal versions flow
- Mentions for agents, channels, and labels to disambiguate a request in the composer
- Saved prompts and conversation history, plus custom context you provide once
- Read-only MCP connector to query your workspace from Claude Desktop, Cursor, or a similar client
Features
- Available from the home page prompt box, a launcher in the top navigation on most screens, and a dedicated full chat view in the sidebar
- Searches and reads across all your conversations, contacts, collected insights, agents, and channels
- Counts and groups conversations by channel, time period, or status, and renders the result as a chart when useful
- Inspects how any agent is configured, including instructions, tone, personality, channels, and comment moderation settings
- Proposes rewrites to agent instructions, tone, and personality, and updates to social comment moderation instructions, each as a before and after comparison you approve or reject
- Approved agent changes are saved as a draft on your agent, so they go live only when you publish that version
- Mention agents, channels, and labels in the composer to make a request unambiguous
- Save prompts you use often, and provide custom context that applies to every conversation
- Long conversations are summarized automatically so you can keep going without losing the thread
- A dedicated settings page at Settings, Ask Dimedove, covering custom context, saved prompts, and credit usage
- Connect Ask Dimedove to an external MCP client to query your workspace from your own tools, with one connector per team and access always scoped to your own workspace
Notes
- Ask Dimedove is available on all plans and consumes credits per request
- The Ask Dimedove MCP connector is read-only and requires authorization
- MCP servers are available to teams on the Pro subscription
Outbound: Reach Your Contacts First
Until now your agents answered. Now they can also reach out. The new Outbound section lets you upload a list of contacts and have an agent call or text every one of them, then carry on the conversation exactly as it would have if the contact had come to you. Upload recipients as CSV or Excel. Only a phone number column is required, and every other column in your file becomes a variable you can drop into your message and that your agent receives as context for that specific person.Outbound calls open with a spoken first message and then run like any other call, including your agent’s configured tasks. Outbound SMS continues as a normal two-way conversation. You can schedule a batch for later in a specific timezone or run it immediately, duplicate it, cancel it, or retry it, and watch per-recipient status as it happens. Every conversation lands in your Inbox alongside all your other channels, linked back to the batch that started it.Reaching out carries obligations, so the guardrails are built in rather than left to you. Scheduling outside 8 AM to 9 PM in the recipient’s timezone requires an explicit acknowledgement, which is recorded on the batch. Recipients can opt out of your SMS at any time with a standard keyword, and opt-outs are honored across SMS batches, call batches, and workflow-triggered outbound alike.
- Outbound calls and SMS from your agents, individually or in batches
- CSV and Excel upload, with every column available as a variable
- Scheduling with timezones, plus duplicate, cancel, and retry
- Live per-recipient status and batch totals, with every conversation in your Inbox
- Calling window guard requiring an explicit acknowledgement outside 8 AM to 9 PM
- SMS opt-out handling honored everywhere your agents reach out
- Outlook Calendar and Applied Epic join the integrations lineup
Features
- A new Outbound section in the dashboard for creating, scheduling, and monitoring batches
- Upload recipients as CSV or Excel, with a downloadable template and a validation preview before you send
- A phone number column is required; every other column becomes a variable usable in your message and passed to your agent as context for that recipient
- Outbound calls open with a spoken first message, with an optional setting to wait for the person to speak first
- Outbound SMS continues as a normal two-way conversation handled by your agent
- Add batch-level context, and post-actions that run after each recipient’s conversation resolves
- Draft, schedule for a specific date, time, and timezone, or run now, and duplicate, cancel, or retry a batch
- Per-recipient status covering queued, calling, answered, voicemail, no answer, sent, opted out, failed, completed, and cancelled, alongside batch totals
- Every recipient conversation appears in your Inbox, linked back to the batch that started it
- Scheduling outside 8 AM to 9 PM in the selected timezone requires an explicit acknowledgement, which is recorded on the batch
- Recipients opt out with STOP, STOPALL, UNSUBSCRIBE, CANCEL, END, QUIT, or OPTOUT, and opt back in with START, UNSTOP, or YES; only a message that is just the keyword counts, so “stop by later” is not an opt-out
- Opt-outs are enforced across SMS batches, call batches, and workflow-triggered outbound, and suppressed recipients are shown as “Opted out”
- Outlook Calendar: connect your Microsoft 365 organization once, choose which mailboxes are bookable, and let your agent check availability, book, and reschedule against real free/busy and Outlook working hours
- Each Outlook host has its own meeting duration, buffers before and after, minimum advance notice, event subject template, and an optional condition telling your agent when to route to them
- Applied Epic: connect the agency management system used by independent insurance agencies, so your agents can look up Clients, Contacts, Companies, Opportunities, Policies, and Accounts, and log Activities against a client
- Integration field mappings can now combine fixed text with values your agent collected, and a new Get operation reads a record without writing to it
- On asynchronous channels, your agent may now stay silent when the last message is a simple acknowledgement and nothing is outstanding, rather than manufacturing a reply
Notes
- Outbound calls and SMS require the Pro subscription with the telephony add-on, and telephony fees are billed separately from any configured monthly limit
- Outlook Calendar and Applied Epic require the Pro subscription with third-party integrations enabled
- Staying silent applies to SMS, Slack, Facebook Messenger, Instagram, and TikTok only, never to voice calls, live chat, forms, or email, and it does not end the conversation
- You remain responsible for consent and for the regulations that apply when you contact people
Credits: A Second Measure of the Work Your Agents Do
Not all agent work looks the same. A conversation that turns into a qualified lead is one kind of outcome. Moderating a hundred comments across your social channels is another: high volume, routine, and genuinely useful, but not something a Qualified Opportunity was ever meant to measure. Credits are the second layer that measures that work.Every plan includes a pool of credits that refreshes each billing cycle, and heavier actions cost more credits than lighter ones, so your usage reflects the real effort involved rather than a flat count. You are never cut off when you reach your allowance: your agents keep working, and anything beyond what your plan includes is billed at your plan rate.Everything is recorded. A new Credits view under Settings shows how much of your allowance you have used, when it resets, a 12-month history, and a full ledger of the actions that consumed credits, so you can always trace usage back to the channels and activities driving it.
- A second usage layer alongside Qualified Opportunities, measuring routine high-volume agent work
- An included pool per billing cycle, refreshing each period
- Weighted by effort, so heavier actions cost more credits than lighter ones
- No interruption at your limit, with usage beyond your allowance billed at your plan rate
- A full ledger of every action that consumed credits
- Inline visibility of what individual actions cost, right where they happened
Features
- Credits sit alongside Qualified Opportunities as a second usage layer and are billed separately
- Each plan includes a pool of credits that refreshes every billing cycle
- Actions are weighted, so more involved work draws down more credits than lighter work
- Your agents keep working past your included credits, with the overage billed at your plan rate
- Included credits do not roll over; each billing cycle starts with your plan’s full allowance
- A Credits view under Settings, Usage shows credits used against credits included, your reset date, and a 12-month history
- A paginated ledger records every action that consumed credits, with its date, activity, and what consumed it
- Individual social comment actions show their credit cost inline in the comment thread view
Notes
- Included credits per billing cycle are 100 on Pay As You Go, 2,000 on Pro, and custom on Enterprise
- Enterprise plans can be configured with a custom credit allowance to match your volume
- If you believe credits were counted incorrectly, contact billing@dimedove.com for review
Channels and Scheduling: TikTok and Live Calendars
TikTok joins your social channels. Connect a TikTok business account, assign an agent, and your agent handles direct messages end to end and moderates the comments on your videos, taking exactly one action per comment: replying publicly, hiding it, unhiding it, or leaving it alone. Instagram caught up in the same release: public replies, hiding, and unhiding are now available there too, where previously only private replies were.Scheduling changed more fundamentally. Calendar tasks used to mean sharing a booking link and hoping the visitor followed it. Now Cal.com, Calendly, and Acuity Scheduling can be read live: your agent checks real availability, books the meeting, and reschedules it without the conversation ever leaving your channel. In chat and forms the visitor picks from an interactive scheduling widget; on voice and SMS your agent offers the same slots as text.
- TikTok as a social channel, with direct messages and comment moderation
- Instagram comment parity, adding public replies, hiding, and unhiding
- Live calendars for Cal.com, Calendly, and Acuity Scheduling
- Availability, booking, and rescheduling inside the conversation, on every channel
- An interactive scheduling widget in chat and forms
- SMS sent from your own team phone number when the conversation is already on one
- Clearer spoken numbers, dates, and confirmation codes on voice calls
Features
- Connect a TikTok business account and assign an agent to handle direct messages and comments
- Your agent takes exactly one action per comment: public reply, hide, unhide, or leave it alone
- Instagram now supports public replies, hiding, and unhiding on comments, alongside private replies
- Every TikTok conversation appears in your unified Inbox with automatic contact creation
- Calendar tasks can run in live mode for Cal.com, Calendly, and Acuity Scheduling, reading real availability rather than sharing a link
- Toggle checking availability, booking, and rescheduling independently for each calendar task
- Configure several event types per task, each with a label, a default, and an optional condition telling your agent when to use it
- Visitors in chat and forms pick a slot from an interactive scheduling widget; voice and SMS conversations get the same slots offered as text
- Agent-sent SMS goes out from the team phone number the conversation is happening on, when that number is SMS enabled
- On voice calls, numbers, currency, dates, times, ordinals, symbols, abbreviations, and units are spoken as full words in the language being spoken
- Phone numbers, postal codes, email addresses, and confirmation numbers are read back one character at a time with pauses, so callers can write them down
Bug Fixes
- A one-off platform hiccup on a comment action no longer flags a social channel as needing attention
- Your “Store call audio” preference is preserved when you publish an agent
- Calendar tasks created before live calendars existed now open in the correct mode
- Adding or changing a phone add-on no longer puts it on its own billing cycle, so your plan and add-ons stay on a single invoice
Notes
- Social channels are available to teams on the Pro subscription
- TikTok has no path from a comment to a direct message, so private replies are not available there, and TikTok public replies are limited to 150 characters
- Like Facebook Messenger and Instagram, TikTok direct messages are inbound only, so your agent does not open the conversation
- Cancelling a booking is deliberately not something your agent can do
Integrations and Reliability: Airtable, MCP OAuth, and Workflows at Scale
Two new ways to connect your agents to the systems you already run on. Airtable joins the record integrations: connect with OAuth and Dimedove discovers every base and table you can write to, along with their fields, so your agents can create, update, and upsert records as they qualify leads and gather information. And MCP servers that require authorization no longer mean managing tokens by hand. Dimedove handles discovery, registration, and refresh for you, and shows you connection status, granted permissions, and expiry at a glance.Voice agents gained transfers, and you configure them by writing them down. Describe your routing rules in plain language anywhere in your agent’s instructions or human fallback text, along the lines of “for billing, transfer to this number; for technical support, route to that one, extension 5”, and they become live transfers on calls. It handles several rules at once, extensions in English and French, and international formatting, and it deliberately does not guess: a number mentioned in passing does not silently become a transfer rule.Workflows also got considerably more dependable under load. Webhook triggers accept an idempotency key so a replay returns the original run instead of starting a duplicate, outbound actions are paced automatically against provider limits with nothing to configure, and a run that gets retried never re-sends a message it already sent.
- Airtable integration with automatic discovery of your bases, tables, and fields
- OAuth support for MCP servers, with automatic token refresh and per-tool control
- Voice call transfers configured in plain language in your agent’s instructions
- Execute Tasks workflow action that picks the right agent task at run time
- Idempotent webhook triggers, automatic pacing, and no duplicate sends on retry
- Agent deletion from the dashboard, with your conversations preserved
- Threaded social comment view showing the full thread, outcome, and reason
Features
- Connect Airtable through OAuth from Settings, Integrations, and Dimedove discovers every base and table you have write access to
- Computed fields such as formulas and rollups are excluded from writes automatically, and you can match on up to three fields when updating existing records
- MCP servers requiring authorization can be connected without managing tokens by hand, with automatic refresh and clear connection status, granted permissions, and expiry
- Where an MCP server needs its own client credentials, the setup screen shows you the redirect URI and takes your client ID and secret
- Re-authorize, request additional permissions, or disconnect an MCP server at any time
- Enable or disable individual tools on an MCP server, which both hides them from your agent and prevents them from being called
- Describe voice call transfer rules in plain language in your agent’s instructions or human fallback text, and they become live transfers on publish
- Transfer rules handle several destinations at once, extensions in English and French, and international number formatting
- A new Execute Tasks workflow action lets you choose an agent, pick which of its tasks are available, add context, and have the workflow decide which task to run and with what inputs
- Send an idempotency key with a webhook trigger and a replay returns the original run instead of starting a duplicate
- Outbound workflow actions are paced automatically against provider limits, with nothing to configure
- Delete an agent from the dashboard with a typed confirmation; live agents must be disabled first, and your conversations are preserved
- Open any social comment to see the whole thread, the reply your agent sent, the moderation outcome, and the reason, with pagination on the activity list
- API tasks support genuinely nested request bodies, including objects inside array items
- Slack agents show a thinking indicator while composing a reply, instead of silence
- The Messenger widget gained a dismiss button on the wide-mode preview that collapses it to the bubble for the session, and formatted text now renders cleanly in the last message preview
- Inbox and Contacts load progressively instead of blocking on a full render, and chat and tab selection live in the URL so deep links and browser navigation work
Bug Fixes
- Agents keep the full results of recent task executions in context, so they no longer forget a list they have just retrieved
- API tasks were sending empty input values
- Agents hold the visitor’s language more consistently through a conversation
- The agent Design form no longer overwrites a saved widget size
- A workflow run that is retried after a transient issue no longer re-sends messages it had already sent
Notes
- Airtable requires the Pro subscription with third-party integrations enabled
- The Messenger widget improvements require SDK version 0.1.7 or later
Apps API: Build Custom Experiences on Dimedove
We are opening up the Dimedove platform to developers. The new Apps API lets you build fully custom chat applications powered by your configured agents. Instead of embedding the Messenger widget, you use a REST API to create conversations, send messages, and receive agent responses in any frontend you design. You own the user experience; Dimedove handles the agentic backend, including agent responses, knowledge base queries, task execution, and conversation context.The API supports real-time streaming (SSE) for responsive chat interfaces and traditional request/response patterns for simpler integrations. Enable Generative UI to let your agent produce interactive form components during conversations, and use tool name mappings to customize how agent tasks are exposed through each App for a clean, developer-friendly API surface.Every App conversation appears in your Dimedove Inbox alongside all other channels, with automatic contact creation and tracking. Built-in metrics and per-app analytics are available directly in the dashboard. To help you get started, we have published an open-source cookbook with a production-ready Next.js chat application.
- Custom chat applications powered by your Dimedove agents through a REST API
- Streaming and non-streaming response modes for real-time or traditional integrations
- Generative UI for agent-generated interactive forms during conversations
- Full dashboard visibility with App conversations in your unified Inbox alongside all other channels
- Tool name mappings to customize how agent tasks are exposed through each App
- Open-source cookbook with a production-ready Next.js chat application template
Features
- Create Apps from the dashboard, assign a published agent, and get your app_id and secret API key
- REST API for creating conversations, sending messages, and managing conversation history with cursor-based pagination
- Server-Sent Events (SSE) streaming for real-time chat, configurable per App or overridden per request
- Non-streaming mode returns complete responses with output arrays for simpler integrations
- Generative UI renders interactive forms through JSON Patch (RFC 6902) operations during conversations
- Tool name mappings let you rename agent tasks per App for a clean, custom API surface
- External user ID support links conversations to your own user system with automatic contact creation
- Built-in analytics with request logs, conversation counts, message volume, and unique user tracking
- Open-source cookbook with a full Next.js chat application demonstrating streaming, Generative UI, and conversation management
Notes
- Apps are available to teams on the Pro subscription
Integrations: Native CRM Support and Instagram
Your agents can now push qualified leads directly into the CRM your sales team already uses. We have rolled out native integrations with six of the most popular CRMs: HubSpot, Attio, Zoho CRM, Pipedrive, Close, and Salesforce. Connect your account through a secure OAuth flow, and your agents can create, update, or upsert records in real time as they qualify leads and gather information during conversations.Each CRM integration automatically discovers your available resources and fields, including custom properties, so your agents always work with the latest schema. Configure agent actions to map conversation outputs to specific CRM fields, choose which resources to target (contacts, leads, deals, companies, opportunities), and select how records should be matched for updates. Salesforce customers can connect using their own Connected App for full control over permissions and security.Alongside CRM support, Instagram is now available as a social channel. Connect your Instagram professional account, assign an agent, and start handling DMs and comment replies automatically. Every Instagram conversation appears in your unified Inbox alongside Messenger, Slack, email, phone, and web chat, giving your team complete visibility across every channel.
- Native CRM integrations with HubSpot, Attio, Zoho CRM, Pipedrive, Close, and Salesforce
- Agent-driven CRM actions to create, update, and upsert records directly from conversations
- Automatic schema discovery that detects your CRM fields, resources, and custom properties
- Instagram social channel with support for DMs and comment replies
- Unified Inbox integration with all CRM-linked conversations and Instagram messages alongside your other channels
- OAuth-based setup for all providers, with Salesforce supporting customer-managed Connected Apps
Features
- Connect any supported CRM provider through a secure OAuth flow from Settings > Integrations
- Automatic resource and field discovery, including custom properties, after connecting
- Configure agent actions to create, update, or upsert records based on conversation data
- Field mapping connects agent outputs to specific CRM fields with lookup support for matching existing records
- Supported CRM resources: contacts, leads, deals, companies, opportunities, and custom objects (varies by provider)
- Salesforce supports customer-managed Connected App authentication with My Domain and PKCE security
- Refresh schemas and re-sync at any time; scope validation prompts reconnection if permissions are missing
- Connect your Instagram professional account and assign an agent to handle DMs and comment replies
- All Instagram and CRM-linked conversations appear in your unified Inbox with automatic contact creation
Notes
- CRM integrations are available to teams on the Pro subscription with third-party integrations enabled
- Instagram is available to teams on the Pro subscription with the Social Channels feature enabled
Socials: Deploy Your Agents to Facebook Messenger and Instagram
Social channels are here. Deploy the same agents you have configured on Dimedove to the social messaging platforms your customers already use. Connect your accounts, assign agents, and start qualifying leads and engaging customers beyond your website. Messenger and Instagram are the first supported channels, with more social platforms coming soon.Connect your Facebook Business Pages or Instagram professional accounts through a secure OAuth flow, then create a social channel from your dashboard. Assign an agent, pick the page or account to connect, and toggle the channel on. Your agent immediately starts responding to incoming Messenger and Instagram direct messages using the same knowledge, instructions, and personality you have configured on the platform. For Facebook Pages, your agent can also moderate comments on your posts according to the business moderation criteria you define, hiding, replying to, or escalating comments automatically. Every conversation appears in your unified Inbox alongside web chat, forms, email, phone, and Slack, giving your team complete visibility across all channels.
- Facebook Messenger and Instagram support as the first social channels, with more platforms coming soon
- Seamless agent deployment to your Facebook Business Pages and Instagram professional accounts in a few clicks
- Comment moderation for Facebook Pages that follows the business moderation criteria you define
- Direct message handling across Messenger and Instagram with full agent context
- Unified Inbox integration with all Messenger and Instagram conversations alongside your other channels
- Automatic contact creation for every Messenger and Instagram visitor
- Flexible channel management to enable, disable, or reassign agents at any time
Features
- Connect your Facebook or Instagram account through a secure OAuth flow directly from the dashboard
- Create and manage social channels with a dedicated management interface
- Select which Facebook Business Page or Instagram professional account to connect to each social channel
- Assign any published agent to handle conversations from a connected page or account
- Toggle channels on and off instantly without losing your configuration
- Agent responds directly in Messenger and Instagram DMs using your configured knowledge base, instructions, and personality
- Define business moderation criteria so your agent can automatically hide, reply to, or escalate comments on your Facebook Pages posts (comment moderation is specific to Facebook Pages and is not available on Instagram)
- All Messenger and Instagram conversations appear in your Dimedove Inbox with full conversation history
- Contact profiles are automatically created for Messenger and Instagram visitors
- Each page or account can be assigned to one social channel for clear routing
- Disconnect individual channels or your entire Facebook or Instagram account at any time
Notes
- Social channels are available to teams on the Pro subscription
Evaluations: Confidence in Every Conversation
Evaluations give you a systematic way to verify that your agent handles real-world scenarios the way you expect. Define a user persona and a set of success criteria, and Dimedove simulates a full multi-turn conversation with your agent, then scores the result with an independent AI judge. Every time you update your agent’s instructions, knowledge base, tone, or tasks, you can run your evaluations to confirm nothing broke and everything still works.Each evaluation follows a three-step lifecycle: define, run, review. The simulated user stays in character throughout the conversation, writing naturally and reacting realistically to your agent’s responses. Once the conversation completes, an independent judge evaluates the full transcript against each criterion you defined, producing a score and a pass, partial, or fail verdict. A detailed reasoning section and per-criterion breakdown show exactly what worked, what did not, and why.Run evaluations individually or all at once after making changes. Every run is saved with its score, verdict, and full conversation transcript, so you can track how your agent’s performance evolves over time. Pair evaluations with Versions to test a draft configuration before publishing it to production.
- Scenario-based testing with realistic, natural user simulation
- Automated scoring with an independent judge evaluating each success criterion
- Pass, partial, and fail verdicts with scoring for granular insight
- Full conversation transcripts for every run, viewable turn by turn
- Run individually or all at once to validate your agent after any change
- Run history to track performance over time and catch regressions
Features
- Create evaluations with custom user scenarios describing who the visitor is and what they are trying to accomplish
- Define success criteria that tell the judge exactly what a passing conversation looks like
- Configure the maximum number of conversation turns per evaluation
- Simulated users stay in character, write naturally, and react realistically to your agent’s responses
- Independent AI judge evaluates the full transcript against each defined criterion
- Scoring system with pass, partial, and fail verdicts based on how well your agent met the defined criteria
- Detailed evaluation reasoning citing specific moments in the conversation
- Per-criterion breakdown showing which criteria passed or failed and why
- Full conversation transcript viewable turn by turn for every completed run
- Run a single evaluation or all active evaluations at once from the dashboard
- Run history with status, date, score, and verdict for every past execution
- Visual status indicators for quick scanning: green for pass, yellow for partial, red for fail
- Evaluations run against your current agent configuration without affecting production conversations, inbox, or contacts
- Bilingual evaluation output in English and French
Notes
- Evaluations are available to teams on the Pro subscription
Workflows: Automate Across Every Channel
Workflows bring event-driven automation to Dimedove. Connect any platform event to a sequence of actions, without writing code or leaving the dashboard. When a form is submitted, an email arrives, a call comes in, an agent qualifies a lead, or a Slack message is received, workflows let you automatically respond with emails, SMS messages, outbound calls, Slack messages, or HTTP requests to external systems.Each workflow starts with a trigger (the event that kicks things off) and executes up to two action steps in sequence. Nine trigger types cover every channel on the platform, and optional filters let you narrow triggers to specific agents, forms, phone numbers, email channels, or Slack agents. Six action types give you full cross-channel reach, including a Delay action for timed sequences like waiting before a follow-up call. When actions are sent through agent-assigned channels, they automatically create conversations in your Inbox, so your team has full visibility and agents can continue the interaction.Template variables make workflows dynamic: reference any field from the trigger payload in your action configurations using
{{trigger.body.field_name}} syntax, with dot notation and array index support for nested data. A built-in run history with step-by-step details and an analytics dashboard give you complete visibility into every execution.- Multiple trigger types covering webhooks, manual runs, qualified opportunities, human assistance requests, form submissions, inbound emails, inbound calls, inbound SMS, and Slack messages
- Multiple action types including HTTP Request, Email, SMS, Call, Slack Message, and Delay
- Optional trigger filters to narrow events to specific agents, forms, phone numbers, email channels, or Slack agents
- Template variables for dynamic data in action configurations
- Inbox integration with automatic conversation creation from email, SMS, and call actions
- Run history and analytics with step-by-step execution details, success rate, and duration tracking
Features
- Create workflows from the dashboard with a visual editor showing triggers, actions, and connections
- Choose from multiple trigger types: Webhook, Manual, Qualified Opportunity, Human Required, Form Submission, Email Received, Inbound Call, Inbound SMS, and Slack Message Received
- Apply optional filters to narrow triggers to specific platform resources
- Webhook triggers generate a unique URL for receiving external HTTP POST requests
- Manual triggers add a Run button to the workflow for on-demand execution
- Configure multiple action nodes per workflow, with Delay supported for timed sequences
- HTTP Request action supports GET, POST, PUT, PATCH, DELETE with configurable headers, query parameters, and JSON body
- Email action with three sending modes: Default (Dimedove), Custom sender name, or through a configured Email Channel
- SMS action from a Dimedove number or a team phone number with agent-assigned inbox conversations
- Call action initiates outbound voice calls through team phone numbers with voice-enabled agents
- Slack Message action posts to any channel in your connected workspace
- Delay action pauses execution for a configurable duration in seconds, minutes, or hours
- Template variables (
{{trigger.body.field}}) in all action text fields with dot notation and array index support - Three workflow statuses: Draft, Live, and Paused
- Run history table with status, trigger type, duration, and step progress for every execution
- Detailed run view showing trigger data, step inputs, outputs, and error messages
- Analytics dashboard with runs over time, total runs, success rate, and average duration
Notes
- Workflows are available to teams on the Pro subscription
Emails: Forward, Respond, Repeat
Emails let your agents handle inbound email conversations. Forward emails from your business addresses to a Dimedove-generated inbound address, assign an agent, and responses go out automatically. You can verify your own domain to send replies from your business address instead of the Dimedove address, with DNS-based verification for DKIM authentication.All email threads appear in your unified Inbox alongside chat, phone, SMS, and Slack. Team members can take over conversations when needed, with configurable handoff behavior: permanent takeover or automatic resume after a timeout. Composition settings include custom signatures, CC/BCC recipients, quoted thread inclusion, and original CC preservation.
- Inbound email forwarding from any business address to your AI agents
- Automatic AI responses using your agent’s full knowledge base and capabilities
- Custom domain verification with DKIM support for professional reply addresses
- Proper email threading with RFC 5322 compliant headers
- Flexible human handoff with permanent or timeout-based resume options
- Unified Inbox integration showing all email conversations alongside other channels
- Configurable composition including signatures, CC/BCC, and quoted threads
Features
- Create emails from the dashboard with automatic inbound address generation
- Forward from any provider that supports mail forwarding (Gmail, Outlook, Yahoo, etc.)
- Assign any published agent to handle email conversations
- Verify custom domains through DNS records for professional reply addresses
- Choose between Dimedove address or verified domain for outgoing replies
- Configure human handoff: permanent takeover or resume after configurable timeout
- Add custom email signatures appended to all agent replies
- Set static CC and BCC recipients for compliance or team visibility
- Toggle quoted thread inclusion and original CC preservation
- All email threads visible in unified Inbox with full conversation history
Notes
- Emails are available to teams on the Pro subscription and above
MCP Support: Extend and Expose Your Agents
We’re introducing full Model Context Protocol (MCP) support, giving your agents two powerful new capabilities: connect to external MCP servers to extend what your agents can do, and deploy your own hosted MCP servers to expose your agents to the broader AI ecosystem. Whether you want to bring third-party tools into your agent conversations or make your agents accessible from ChatGPT, automation platforms, and other AI tools, MCP makes it possible.MCP Tasks let your agents connect to any MCP-compatible server during conversations. Configure the server URL, choose your transport protocol, add custom headers for authentication, and your agent automatically discovers and uses the available tools. This opens up integrations with specialized services, custom APIs, and external capabilities that extend your agent far beyond its built-in functionality. A trust acknowledgment ensures you maintain control over which third-party servers your agents can access.MCP Servers flip the script: instead of consuming external tools, you expose your Dimedove agents as tools that other applications can use. Create a hosted MCP server from your dashboard, assign an agent, select which tasks to expose, and optionally enable knowledge base access. Dimedove manages all the infrastructure, so there are no servers to deploy or maintain. Your agents become accessible to ChatGPT apps, AI automation tools like n8n, Make, Gumloop, and Zapier, AI assistants and copilots, and any MCP-compatible client. Built-in rate limiting and analytics give you visibility and control over external usage.
- MCP Tasks for connecting agents to external MCP servers and third-party tools
- MCP Servers for exposing your agents via Dimedove-hosted MCP endpoints
- Selective task exposure to control exactly what external tools can access
- Knowledge base access toggle for MCP server queries
- Connection testing to verify MCP configurations before going live
- Built-in rate limiting and usage analytics for MCP servers
Features
- Create MCP tasks that connect to any external MCP server during conversations
- Configure server URL, transport protocol (Streamable HTTP or SSE), and custom headers
- Automatic tool discovery from connected MCP servers
- Trust acknowledgment workflow for third-party MCP server connections
- Create and manage MCP servers from Settings > MCP in your dashboard
- Assign any published agent to an MCP server
- Selectively expose specific agent tasks through each MCP server
- Enable or disable knowledge base queries per MCP server
- Test MCP server connections and view discovered tools before enabling
- Copy MCP server URLs for easy configuration in external tools
- Live/Draft status with validation ensuring proper configuration before enabling
- Per-server rate limiting to prevent abuse and ensure reliable performance
- All MCP tool calls logged for monitoring and debugging
Notes
- MCP Tasks and MCP Servers are available to teams on the Pro subscription
Multiple Agents: Specialized Experts for Every Channel
Teams can now create and manage multiple agents, each becoming an expert in a specific part of your lead qualification flow. Dimedove makes it seamless to assign agents to Forms, Slack channels, Telephony, and more channels coming soon, so you can have specialized qualification agents tailored to each channel’s unique context and audience.Duplicating an existing agent is now as simple as a right-click. All your carefully configured settings, tasks, and prompts are copied to the new agent, letting you build on proven configurations without starting from scratch. Your agent’s content (knowledge base files and websites) starts fresh, giving you full control over what each agent knows.
- Create multiple agents based on your team’s plan
- Duplicate existing agents with all configuration, tasks, expertise, and prompts copied
- Assign agents to channels including Forms, Telephony, and Slack
- Specialize qualification flows with agents tailored to each channel
- Manage limits through your team’s billing settings
Features
- Create new agents directly from the dashboard with a simple name input
- Duplicate any existing agent to create a new agent with the same configuration
- All behavior, design, language, task, expertise, and prompt settings are copied during duplication
- Agent content (files, websites) and preview conversations start fresh for each new agent
- Assign different agents to Forms, phone numbers, and Slack channels
- View and manage agent limits in team billing settings
- Each agent operates independently with its own knowledge base and conversation history
Slack Agents: Qualify Leads Where Work Already Happens
We’re excited to introduce Slack Agents, a powerful new channel that brings your Dimedove agents directly into your Slack workspace. Qualify leads right where your team already works, with consistent agent responses that keep everyone aligned.Connect your Slack workspace through a secure OAuth flow, then create Slack Agents and deploy them to public and private channels. Choose whether your agent responds only when @mentioned (recommended for busy channels) or participates in all thread conversations. All responses are organized in threads, keeping your channels clean while enabling rich, contextual discussions.Every Slack conversation appears in your unified Dimedove Inbox, giving you complete visibility across all channels. Whether a conversation happens on your website, through a form, over the phone, or in Slack, you have one place to see it all.
- Slack workspace connection via secure OAuth integration
- Deploy to public and private channels with simple slash commands
- Flexible response triggers: respond on @mention only or to all thread messages
- Threaded conversations for organized channel management
- Unified Inbox integration showing all Slack conversations alongside other channels
Features
- Connect your Slack workspace to Dimedove through Settings > Integrations
- Create multiple Slack Agents for different purposes or teams
- Assign any published Dimedove agent to handle Slack conversations
- Configure response triggers: mention-only (default, recommended) or all messages in threads
- Add agents to channels using
/invite @Dimedoveand/dimedove add {agent_id}commands - Remove agents from channels with
/dimedove remove - All agent responses posted as threaded replies for clean channel organization
- View and manage all Slack conversations in the Dimedove Inbox
- Enable/disable Slack Agents from the dashboard without removing channel assignments
Notes
- Slack Agents are available to teams on the Pro subscription
- Currently available in public and private channels only (direct message support coming soon)
Attachments & Dictation: Richer Conversations
We’re introducing two powerful new ways for visitors to communicate with your agents: file attachments and voice dictation. These features make conversations richer, more accessible, and more efficient for everyone.With attachments enabled, visitors can share images and documents directly in the chat. Your agent can analyze uploaded files, extract relevant information, and use that context to provide better responses. No more asking visitors to copy and paste document contents or describe images in detail.Dictation brings voice input to your agent conversations. Visitors can tap the microphone button and speak their messages naturally. Their speech is transcribed in real-time using their device’s built-in recognition, and they can review and edit the text before sending. This is especially valuable for mobile users and makes your agent more accessible to everyone.
- File attachments for sharing images and documents in conversations
- Document analysis enabling agents to extract and reference information from uploaded files
- Voice dictation using device-native speech recognition
- Hands-free input ideal for mobile users and accessibility
- Real-time transcription with editing before sending
Features
- Enable attachments in the Design tab to allow visitors to upload files
- Support for most popular image formats (screenshots, photos, visual content)
- Support for common document formats (PDFs, spreadsheets, presentations, text files)
- Agents automatically analyze uploaded files and extract relevant information
- Enable dictation in the Design tab to add voice input to conversations
- Microphone button appears in chat interface when dictation is enabled
- Speech transcribed in real-time using device’s native speech recognition
- Recording automatically stops when visitor finishes speaking
- Transcribed text appears in input field for review and editing before sending
- Dictation language matches agent’s default language setting
- Works on web browsers and mobile devices that support speech recognition
Notes
- Attachments are available to teams on the Pro subscription
- Dictation is available to all teams
Agent Versions, Labels, and Forms Upgrades
We’re rolling out a major set of improvements that help teams iterate faster, organize conversations at scale, and capture data more reliably. First up: Agent Versions. You can now create multiple versions of an agent’s configuration, preview how each behaves, and switch or roll out changes with confidence using a two-step deployment flow. This makes it easy to A/B test instructions, tasks, or knowledge and evolve your agent without disrupting production.Next, we’re introducing Labels: a powerful way to automatically categorize conversations. Define the labels that matter to your business (e.g., product fit, priority, intent), and Dimedove will auto-assign them based on the exchange. Labels appear right in the Inbox for quick scanning and can be managed centrally, giving your team new ways to segment, personalize, and report on outcomes.We’ve also upgraded Forms to be more resilient and structured. You can enable auto-submit to finalize a Form as soon as the agent collects all required inputs, and specify the exact question order the agent should follow. Under the hood, we improved how agents handle task inputs, especially AI-inferred prompts, and tightened email format validation for cleaner data.To round things out, we added a conversation toolbar in the Inbox so you can mark threads as requiring human assistance, quickly export messages, and manually assign labels when needed. A new notifications center lets each teammate subscribe to essential email alerts like “human assistance required” or “qualified opportunity,” so nothing slips through the cracks.And on mobile, the Messenger received a significant UX upgrade with much smoother virtual keyboard handling and layout polish to keep conversations fluid on every device.
- Agent Versions to safely test, preview, and deploy configuration changes with a two-step flow
- Auto-assigned Labels for smarter conversation categorization and faster prioritization in the Inbox
- Inbox conversation toolbar to request human help, export messages, and assign labels manually
- Forms upgrades with auto-submit, precise question ordering, and stronger task/input handling
- Notifications center for opt-in email alerts on key events
- Mobile Messenger improvements with drastically better virtual keyboard handling and UX on small screens
Features
- Create and manage multiple agent configuration versions; preview outputs, A/B test behavior, and switch live versions without downtime
- Two-step deployment separates editing from publishing so production remains stable while you iterate
- Define team labels that are automatically assigned to conversations based on agent understanding; view and filter labels directly in the Inbox
- Manage labels in the dashboard with plan-aware caps for PAYG; labels also appear inline in conversation views
- New Inbox conversation toolbar: mark conversations as requiring human assistance, export messages in one click, and manually assign labels
- Forms: enable auto-submit to complete a Form as soon as required data is captured; set the exact order of inputs for a consistent question flow
- Improved Form task handling and required AI-inferred input collection for more reliable data capture; stricter email format validation for cleaner records
- Notifications center (baseline): subscribe to key email alerts such as “human assistance required” and “qualified opportunity”; expand deeper via webhooks
- Admin quality-of-life: added search filters across admin views to find what you need faster
- Messenger: mobile UX revamp with better virtual keyboard handling, improved conversation starter alignment, optional hide-close in compact mode, and automatic light/dark mode fallback
Bug Fixes
- Fixed a dialog cleanup issue that could block screen interactivity after closing
- Resolved grouping inaccuracies in usage charts for performance records
- Improved mobile Messenger initialization for conversation starters and device theme detection
- Corrected compact mode behavior with an option to hide the close button where desired
- Addressed various mobile layout and virtual keyboard edge cases for a smoother typing experience
- Improved Form behavior when overriding team tasks and ensured email formatting is respected during collection
- Refined dashboard data refreshes for webhooks, Forms, and phone resources to reduce stale views
Messenger Compact Mode: Branding, Visibility, and Mobile Controls
We’re rolling out powerful new customization options for the Dimedove Messenger, especially in compact mode. You can now add your company logo, fine-tune exactly what’s displayed on desktop and mobile, and guide visitors with streamlined quick replies. Conversation Starters and Smart Replies are unified into tappable suggestions, making it faster for prospects to jump into the right flow. Everything is configurable in the Design tab and updates instantly with no code required.
- Customizable compact mode with company logo branding
- Per-device visibility controls to tailor the Messenger on desktop and mobile
- Unified quick replies combining Conversation Starters and Smart Replies
- Seamless embedding with improved width behavior and non‑blocking layout
Features
- Configure Messenger compact mode directly in the Design tab
- Upload and display your company logo to match your brand
- Control what appears on desktop vs. mobile (e.g., headers, subtitles, quick replies, conversation starters)
- Conversation Starters and Smart Replies now render as consistent quick replies for faster interactions
- Safer preview experience with graceful handling of invalid preview links
- Updated browser tab titles to better reflect your brand and context
Bug Fixes
- Fixed compact mode width handling and removed dead space that could block page content
- Improved chat error handling and automatic redirect behavior at the end of Form flows
- Minor visual fixes: improved link colors in light mode and clearer close button styling
- Telephony polish: smoother phone/SMS UX with performance optimizations, more natural voice pacing during tool use, and improved multilingual voice handling
- More reliable external ID capture in Forms and an in-dashboard webhook payload preview for easier testing
- General stability and performance improvements across Messenger and chat
Conversational Forms: Human-Like Data Collection at Scale
We’re excited to introduce Forms, a revolutionary approach to prospect qualification that transforms rigid data collection into natural, engaging conversations. Traditional forms force prospects into checkboxes and text fields, limiting the richness and authenticity of the information they share. Forms solve this by deploying your AI agents in a dedicated channel explicitly designed for conversational data collection, enabling you to gather diverse, nuanced insights that traditional input fields could never capture.Forms are perfect for businesses qualifying prospects at scale who need specific datapoints to optimize their conversion flow. Each Form gives your agent a focused mission: collect a defined set of inputs through natural dialogue while leveraging all your existing agent configurations and instructions. You can customize the agent’s behavior specifically for each Form, ensuring the conversation aligns perfectly with your qualification goals. The beauty is in the flexibility: prospects can share complex information, clarify details, and provide context in their own words, all while your agent intelligently guides them through the process.Setting up Forms is effortless. Create a Form, define the data you need, customize your agent’s approach, and Dimedove hosts and manages the entire experience. Each Form gets a shareable link you can embed anywhere: in emails, on landing pages, in ads, or wherever your prospects are. You can assign external IDs for tracking, configure redirect URLs for post-submission flows, and set up webhooks to instantly push collected data into your own systems. Every Form conversation appears in your unified Inbox alongside chat, phone, and SMS interactions, giving you complete visibility across all engagement channels.Forms also support full internationalization, adapting to any language your agent’s business supports, and include built-in preview and testing capabilities so you can perfect the flow before going live.
- Conversational Forms for natural, human-like data collection
- Rich data capture beyond traditional form inputs and constraints
- Agent-powered qualification at scale with intelligent conversation flows
- Fully hosted and managed shareable Form links by Dimedove
- Custom Form configurations with agent behavior tailored per Form
- External ID tracking and post-conversation webhooks for seamless integrations
- Unified Inbox integration showing all Form conversations alongside other channels
- Multi-language support with automatic Form internationalization
- Preview and testing capabilities to validate flows before deployment
Features
- Create unlimited conversational Forms for prospect qualification and data collection
- AI agents collect defined datapoints through natural, engaging dialogue
- Forms leverage existing agent configurations with per-Form customization options
- Fully hosted Form links managed by Dimedove for easy sharing and embedding
- Assign external IDs to Forms for tracking and CRM integration
- Configure redirect URLs to control post-submission user flows
- Post-conversation webhooks to export collected data to external systems in real-time
- All Form conversations stored in unified dashboard Inbox with chat, phone, and SMS
- Rich data collection beyond traditional form inputs: context, clarifications, and nuanced responses
- Built-in Form preview and testing mode to validate conversation flows
- Automatic internationalization supporting all agent-configured languages
- Form-specific agent instruction customization for optimized qualification flows
- Shareable Form links embeddable in emails, landing pages, ads, and campaigns
Bug Fixes
- Improved Form webhook delivery reliability for high-volume scenarios
- Fixed edge cases in multi-step Form conversation state management
- Enhanced Form redirect URL handling for complex parameter passing
- Resolved Form preview mode reset issues after configuration changes
- Optimized Form loading performance for faster initial conversation startup
Telephony Support: Voice Calls & SMS
We’re thrilled to announce native telephony support, bringing your AI agents to voice and SMS channels. Phone calls and text messaging remain natural, trusted, and accessible ways for many prospects to reach businesses. Whether someone prefers the convenience of a quick text or the personal touch of a phone conversation, your agents are now ready to engage meaningfully on both channels.Setting up telephony is remarkably simple. Purchase a managed phone number directly from the Dimedove dashboard, enable calls and/or SMS, assign your agent, and you’re live. No complex integrations, no technical overhead: just seamless deployment to voice and text channels. Your agents leverage state-of-the-art voice models engineered for low latency and natural, delightful voice experiences that feel genuinely human.What makes this truly powerful is continuity. Every phone call and SMS conversation appears directly in your unified Inbox alongside all other interactions. Insights, contact details, and conversation history flow seamlessly across channels. Even better, visitors are recognized by their phone number, allowing agents to pick up conversations exactly where they left off, whether the last interaction was a text, a call, or a web chat. This level of personalization across voice and SMS creates an agent experience unlike anything available today.
- Native voice and SMS support for AI agents on phone channels
- Managed phone numbers purchasable directly from the dashboard
- Zero-integration setup with instant channel activation
- State-of-the-art voice models for low-latency, natural conversations
- Unified inbox showing all calls, texts, and web conversations together
- Cross-channel recognition with phone number-based visitor identification
- Seamless conversation continuity across voice, SMS, and web channels
Features
- Purchase and manage phone numbers directly from the Dimedove dashboard
- Enable voice calls and/or SMS messaging for any agent with one click
- Assign agents to phone numbers for instant telephony deployment
- Low-latency voice conversations powered by state-of-the-art voice models
- Natural, human-like voice experiences designed for customer engagement
- All phone and SMS conversations accessible in the unified dashboard Inbox
- Automatic visitor recognition through phone numbers across all channels
- Conversation history and context preserved when switching between voice, SMS, and web
- Contact insights and attributes automatically captured from phone interactions
- Separate usage-based charges for telephony services
Bug Fixes
- Improved voice model response latency for faster call interactions
- Enhanced SMS delivery reliability for high-volume scenarios
- Fixed edge cases in cross-channel conversation threading
- Resolved phone number validation issues during purchase flow
- Optimized contact matching logic for phone number recognition
Full Screen Mode & Performance Enhancements
We’re excited to introduce full screen mode, a powerful new deployment option that transforms how you can present your AI agents. This immersive, full-width experience is perfect for businesses wanting to dedicate entire pages to their conversation interface, whether as a standalone support portal, a dedicated consultation page, or an interactive product guide.Full screen mode removes all traditional widget constraints, giving your agent conversations the space they deserve. Your customers get a distraction-free, professional interface that feels like a native part of your brand experience. The setup is effortless: simply navigate to your agent’s Design tab and toggle full screen mode on. Changes apply instantly, and you can switch between full screen and embedded modes anytime without any code changes.This release also includes significant performance improvements across the platform, making your agents faster and more reliable than ever.
- Full screen mode for immersive, full-width agent chat experiences
- Real-time design switching between full screen and embedded modes
- Zero-code deployment with instant configuration updates
- Performance optimizations for faster agent responses
- Enhanced reliability with prompt caching and initialization improvements
Features
- Full screen mode available in the agent Design tab for dedicated conversation pages
- Instant toggle between full screen and embedded widget modes without redeployment
- Optimized full-width interface for immersive customer experiences
- Seamless brand integration with full design control in full screen mode
- Agent prompt caching optimizations for faster response times
- Accelerated agent initialization reducing time-to-first-message
Bug Fixes
- Implemented agent prompt caching optimizations for improved performance
- Fixed duplicate contact creation issue in high-traffic scenarios
- Accelerated agent initialization for faster conversation startup
- Resolved edge cases in full screen mode responsive behavior
- Improved session handling for seamless mode switching
Data Privacy & Custom Agent Performance Metrics
We’re strengthening our commitment to data privacy and security with the implementation of advanced de-identification and anonymization capabilities for third-party LLM processing. This ensures your sensitive data remains protected while maintaining the full power of AI processing capabilities.Additionally, we’re introducing custom agent performance definitions that give businesses complete control over how their agents are measured. Choose from three performance metric options - instructions alignment, task completion, and datapoints gathering - to define exactly how your agents should be incentivized and evaluated.
- De-identification and anonymization for third-party LLM processing
- Custom agent performance metrics with flexible measurement options
- Enhanced data privacy and security commitment
- Improved agent evaluation system for better business alignment
- Minor bug fixes and improvements across the platform
Features
- Implemented de-identification and anonymization for all third-party LLM processing
- Enhanced data privacy and security protocols to protect sensitive information
- Custom agent performance definition system with three metric options: instructions alignment, task completion, and datapoints gathering
- Flexible performance measurement allowing businesses to choose how their agents are incentivized
- Improved agent evaluation system for better business goal alignment
- Comprehensive documentation available at docs.dimedove.com/platform/team/usage
Bug Fixes
- Resolved minor UI inconsistencies in agent performance dashboard
- Fixed data processing edge cases in anonymization pipeline
- Improved error handling for custom performance metric configurations
- Enhanced stability of third-party LLM integration workflows
- Various performance optimizations and minor improvements
GPT-5 Integration + Agent Architecture Overhaul
We’ve significantly improved our mixture of agents architecture powering agent conversations. Dimedove now leverages breakthroughs from the newly released GPT-5 model from OpenAI, including enhancements in dynamic reasoning (think when needed), optimized instruction following, tool calling, and memory/context management. At the core of our agent architecture, we’ve fundamentally enhanced how agents manage conversation history, weigh user memories, and balance chat history with recent messages.This release delivers smarter, more contextual conversations with improved file handling and clearer progress tracking giving your agents deeper understanding and better user experiences.
- GPT-5 model integration with dynamic reasoning and enhanced instruction following
- Overhauled agent architecture for improved conversation history and memory management
- Enhanced loading states with detailed progress tracking for agent operations
- Larger file upload support for richer agent content and interactions
- Improved contact extraction from insights for third-party scenario handling
Features
- Integration with OpenAI’s GPT-5 model featuring dynamic reasoning capabilities
- Enhanced instruction following and advanced tool calling functionality
- Improved memory and context management for more coherent conversations
- Redesigned agent architecture for better conversation history weighting
- Enhanced user memory consideration in response generation
- Detailed loading states showing agent progress: knowledge base search, reasoning, web browsing, contact updates, and response composition
- Expanded file upload capacity for agent content processing
Bug Fixes
- Enabled larger file uploads for comprehensive agent content processing
- Improved contact attributes extraction from insights for third-party scenarios
- Fixed conversation history balance between recent messages and stored context
- Resolved loading state inconsistencies during agent operations
Webhooks + Team Key Management
We’ve rolled out secure webhook signing to help teams verify the authenticity and origin of every request. All outgoing payloads now include signed headers that can be validated with your team’s secret key. Alongside this, we’ve introduced publishable and secret API keys, enabling more secure integrations and scoped installations across environments.This release provides the foundation for building automated workflows that are safe, scalable, and trustworthy, giving your agents more ways to work for you without compromising security.
- Webhook signature headers for tamper detection and origin verification
- Publishable + secret keys for teams managing agents and integrations
- Scoped messenger installs with key-bound validation
- Improved installation security and agent embedding workflows
Features
- Signed webhook headers using team-specific secrets
- Team-level publishable and secret keys now available in dashboard
- Scoped messenger install flows using API key validation
- Install security improvements with tighter key-origin linking
Bug Fixes
- Resolved delayed key regeneration timing bug
- Fixed occasional webhook replay issue on retries
- Corrected key copy-to-clipboard behavior in dashboard
Messenger Package + Memory Updates
The Dimedove Messenger is now live and installable via
@dimedove/messenger, our official NPM package. With a single line of code, you can add our agent UI to any site, from Webflow and Shopify to custom frameworks like Next.js. We’ve also rolled out visual and behavioral enhancements: adaptive smart replies, agent memory improvements, and full security control for domains and CORS.This update empowers teams to launch secure, beautiful, and responsive agent experiences faster than ever.- Messenger install package (
@dimedove/messenger) for frictionless setup - Domain whitelisting and dynamic CORS for secure embedding
- Memory improvements for better conversation history and continuity
- Messenger UX enhancements: translucent UI, embedded feel, auto-adaptive dark mode, design control
- Smart reply system for adaptive conversation suggestions
Features
- Initial release of
@dimedove/messengerpackage - Built-in support for Webflow, Shopify, Wix, and custom frameworks
- Domain restrictions and dynamic CORS handling per workspace
- Messenger design refresh with liquid-glass styling and branding controls
- Smart reply suggestions and improved context-following behavior
- Agent memory improvements for returning visitors
Bug Fixes
- Fixed cross-origin embed inconsistencies with Webflow sites
- Corrected message duplication on fast reply inputs
- Improved initial load performance on low-bandwidth connections
- Resolved scroll locking bug on mobile messenger window
Beta Launch + Infrastructure Hardening
We’ve officially entered beta. This release introduces the full Dimedove dashboard experience, complete with team roles, agent creation and configuration, knowledge source management, and core security controls. Contacts and Inbox are now available to help businesses track interactions, extract structured insights, and manage conversations live. Behind the scenes, we’ve laid the groundwork with hardened cloud infrastructure and improved deployment automation to support the scale ahead.
- Dashboard beta with team roles, agent creation, and knowledge config
- Inbox launch with live view, agent actions, and flagged conversations
- Contacts system with attribute management and automatic insight capture
- Security controls including role-based access, logouts, and domain restrictions
- UX improvements to agent setup, knowledge inputs, and team workflows
- Infrastructure hardening for safer deployments and improved reliability
Features
- Beta dashboard with agent configuration, team access, and usage monitoring
- Contacts launched with insight extraction and custom attributes
- Inbox with real-time visibility, flagged conversation tracking, and summaries
- Role-based access control with admin/member support
- Live agent security settings including allowed domains and sign-out tools
- Deployment pipeline hardened with automated scaling and geo-protection
Bug Fixes
- Resolved inconsistent agent saving states in creation flow
- Fixed contact list sync delay after new conversations
- Corrected edge case crashes on empty team dashboards
- Minor visual fixes in the Inbox session viewer