Release notes

The Bulk End-User Deletion API is now available. It erases all personal data associated with one or more end users across Ada’s systems, supporting compliance with data deletion requirements such as GDPR.

API endpoints

Two REST endpoints are available:

• Submit a deletion request (POST /v2/end-users/delete): Accepts up to 1000 identifiers and returns a job_id for tracking. Requests are idempotent for 24 hours.
• Get a deletion job (GET /v2/jobs/{job_id}): Returns the job status and a per-identifier outcome for each submitted identifier.

The v2 Conversations endpoint of the Data Export API now returns two additional fields on each conversation:

• last_agent_id — the unique ID of the most recently assigned Agent
• last_agent_name — the Agent’s name, corresponding to last_agent_id

Both fields are the final entry of the existing agent_id and agent_name arrays (chronological by first-join). They return null when no Agent was assigned to the conversation.

The email footer character limit has been increased from 2000 to 2200 characters, giving more room for legal disclaimers, compliance messaging, and localized content.

Learn more in Email appearance settings.

The MCP Server page (Platform > MCP Server) now lists one connection per AI assistant application. If a person has connected more than one assistant (for example, Claude and Cursor), each appears as its own row showing the application, the authorization date, and the person’s email.

Revoking a connection now removes access for that application only. The person’s other connected assistants keep their access. Previously, revoking removed the person’s access across all of their connected assistants at once.

The Ada dashboard now has an MCP Server page under Platform. It lists every AI assistant connection authorized for your instance — the email of the person who connected and the date they connected — so you can see who has access. Admins can revoke a connection to stop that AI assistant from using the MCP Server; to regain access, the person must authorize again.

Creating, updating, and deleting custom Glossary terms is now recorded in the Audit log. Each entry captures who made the change, the term affected, and the interface it came through — the dashboard or Ada’s MCP server.

Learn more about the Audit log.

The Conversations endpoint in the Data Export API (v2) now returns agent_id and agent_name in the order each Agent joined the conversation. Position [0] is the first Agent assigned; the last position is the most recent.

The two arrays remain paired — agent_name[i] corresponds to agent_id[i] — and the order is stable across repeated calls.

Team member roles can now be changed directly from the Team page. The Role column shows a dropdown for each member you have permission to edit, and selecting a new role saves immediately.

Previously, changing a role required removing the team member and re-inviting them with the new permission level.

Who it affects

On social channels, media and other non-text messages now go to your AI Agent instead of triggering a fixed reply — so the Agent responds in context and explains what it can help with.

What’s changed

• Previously, when a customer sent an image, video, audio, or other file your AI Agent couldn’t read on a social channel (such as WhatsApp), the Agent returned a fixed “Unsupported Messages” response (for example, “Oops! I can only read text”).
• Now, those messages are passed to your AI Agent, which replies directly rather than sending the static response.
• The Handling media messages section, including the editable Unsupported Messages response, has been removed from Config > Channels > Social.

Glossary lets you teach an AI Agent your business’s vocabulary. You provide the terms your end users use and the corresponding term you want said back. This feature ensures the Agent stays consistent and accurate across every language it supports and within a single language. It shapes both what the Agent says and what it looks up, including Knowledge, Playbooks, and other entities. You manage it entirely in the dashboard, including bulk CSV import.

The Knowledge API now supports availability rules on articles. An availability rule controls which articles are available to the AI Agent during a conversation, based on variables like language, user role, or any custom metadata.

A new availability_rules field is supported on:

• GET /v2/knowledge/articles/ and GET /v2/knowledge/articles/{id}
• POST /v2/knowledge/bulk/articles/

You can now control, for each piece of information your Voice AI Agent collects, how callers provide it and whether the Agent reads it back to confirm — so critical values like order numbers get verified, while quick, low-stakes values stay fast.

What’s new

• Capture settings — let callers respond by speech, phone keypad, or text message, or combinations, depending on what you’re collecting.
• Confirmation settings — choose how spoken values are confirmed: no read-back, a natural repeat, a full read-back-and-confirm, or a character-by-character read-back for alphanumeric values.
• Available for both Action inputs and Playbooks Ask steps.

The Persona API is now generally available. It provides programmatic read and update access to an AI Agent’s persona configuration (the settings that shape how the Agent communicates found under Config > Settings > Preferences) so persona changes can be managed through code and version control instead of only through the dashboard.

API endpoints

The Conversations endpoint in the Data Export API (v2) now exposes two new fields:

• used_articles: the knowledge articles referenced during the conversation, each with id, name, url, and source.
• used_playbooks: the Playbooks invoked during the conversation, each with id, name, playbook_execution_id, and outcome_status.

Voice Simulations now run through the same voice pipeline as live calls, so what you hear in a simulation is what your callers experience in real life.

They now cover behavior that only happens on a real call—speech recognition, natural pacing, and real-world latency—and each result includes a recording of the call alongside the transcript and pass/fail evaluation.

A record of the configuration changes across your instance, who changed what, from where, and when, can now be pulled into your SIEM or other security tooling for monitoring and reporting.

Retrieve the Audit log on demand through the API, or subscribe to webhook events to receive change activities in real time.

Validation errors from the v2 API now return details in a single, predictable shape. Previously this field could appear as an index-keyed object, a list of error objects, a bare string, or null, which made it hard to parse reliably.

For validation errors, details is now always a list of error objects (each with parameter, location, and message), or null when there is no field-level detail:

The MCP Server now supports Playbooks as an entity type across both list_entities and propose_change.

Read Playbooks — Use list_entities with entity_type: "playbooks" to list all Playbooks, or pass an entity_id to fetch a single Playbook with full step detail.

Author Playbooks — Use propose_change to create, update, enable, and disable Playbooks directly — including full step authoring with send, ask, set, run, if_else, and go_to step types. New Playbooks are created inactive by default. Active Playbooks can be updated in place or duplicated to apply changes to a copy.

The Conversations endpoint in the Data Export API (v2) now exposes a new classifications field.

The field returns the Topic and Intent classifications for a conversation. Each entry includes a topic_id, topic_name, and a list of intents, where each intent includes intent_id, intent_name, and status. Conversations without classifications return an empty array.

Intent Intelligence is now generally available. Conversation Topics now break down into Intents — a more granular classification layer within each Topic — for a more precise view of why end users reach out.

Topics and Intents

Each Topic groups conversations by subject and contains Intents that capture the specific reason behind a conversation. The Analytics > Topics & Intents page shows volume, AR rate, AR opportunity, CSAT, and containment rate for every Topic and Intent in an expandable table.