Tactiq
by Tactiq
Search meetings and read transcripts from Google Meet, Zoom and Teams inside your AI assistant. 9 tools, four meeting-scoped OAuth permissions verified live, and a staged excerpt-then-expand design for transcript content.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Connect Tactiq via MCP
https://mcp.tactiq.io/Works in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.
Tactiq Tools & Capabilities (9)
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
Limits
- Transcript reading needs a Team seat. Free and Pro accounts can search and list meetings only. Tactiq states the assistant cannot read summary content on those plans.
- Tactiq's own pages disagree on which plan gates MCP. The help centre says everyone can connect; the pricing page lists Tactiq MCP (beta) as a Business-plan feature alongside a separate Tactiq Claude Connector (beta). We could not reconcile these, and the beta label appears only on pricing.
- Permissions do not upgrade themselves. Tactiq states a connection keeps the permissions approved when it was created, so upgrading your plan later does not grant the details permission. You must disconnect and reconnect.
- The staged excerpt design is an affordance, not an enforced limit. All three transcript tools sit behind one permission; nothing prevents a client calling
get_transcriptdirectly. - We could not read tool schemas or safety annotations. The endpoint is gated, so the tool count and names are Anthropic's directory snapshot and no tool here is annotated. What we verified live was the auth posture, both OAuth descriptors and the four scopes.
- No prompts or resources were readable. Anthropic's snapshot records zero prompts, and the gated endpoint meant we could not check whether the server declares either capability.
- The no-training commitment covers the model provider, not Tactiq. It describes Azure OpenAI's handling of API data. We found no Tactiq page committing that Tactiq itself never trains on transcripts.
- Participants are not notified by default. The in-meeting notice is an opt-in setting, and disabling the transcript notification is sold as a Pro-plan feature.
- AI credits are metered on lower plans. Tactiq states Free includes 5 AI credits a month and Pro 10, that one credit covers a single AI action such as generating a summary, and that AI features pause when the limit is reached while transcription continues. Team and above are unlimited.
- Archived meetings disappear after 30 days. Anything the connector could read yesterday may be permanently gone today if it was archived.
- This page describes data access only. It is not legal advice on recording consent.
Frequently asked questions
Is the Tactiq MCP connector read-only?
Yes, on every source we checked. All nine tool names read — list, search, get and expand — and Tactiq's own documentation describes only reading. The four OAuth permissions the server requests are meeting-scoped rather than action-scoped, and none names a write. Anthropic's directory publishes no permissions label for Tactiq at all, so the read-only finding is ours rather than a repeated label.
What does the Tactiq MCP server let an AI assistant read?
Meeting metadata on any plan, and transcript content only on a Team plan. Free and Pro accounts can search and list meetings. Team or higher adds meeting details, AI summaries, generated artifacts and full transcription. The split is enforced by a separate OAuth permission, so the transcript half can be declined at the authorisation screen even on a Team plan.
Does Tactiq join my meeting as a visible bot?
No. Tactiq transcribes inside the browser through a Chrome or Edge extension rather than joining the call as a participant, so no bot appears in the attendee list. Tactiq states audio is not stored and only text transcripts are kept. Other attendees see nothing unless you turn on the optional in-meeting notification.
Are other meeting participants told that Tactiq is transcribing?
Only if you switch it on. Tactiq's notification is a setting, not a default: enabling it posts a one-time message in the meeting chat at the start of each call. Tactiq also states the live transcript display is visible only to you, and its Pro plan sells the ability to disable the transcript notification entirely.
Which meetings can one Tactiq MCP authorisation reach?
Whichever of the four permission groups you approve. The permissions split your own meetings, meetings shared with you and meetings in shared Spaces into separate grants, so a colleague's shared recording is reachable only if you approve that specific scope. Tactiq states the same access rules as the Tactiq app apply — the connector adds no new access.
Does Tactiq use my meeting transcripts to train AI models?
Tactiq states its model provider does not train on the data. It routes AI generation through Microsoft Azure OpenAI's enterprise API, says content logging is disabled by a service modification, and says data sent that way is not used to train or improve those models. That commitment describes the provider's pipeline, not Tactiq's own retention, and Tactiq's other pages name plain OpenAI instead.
Why does Claude say the mcp:meetings:details permission is required?
Your existing connection was authorised without it. Tactiq states connections keep whatever permissions were approved when created, so upgrading to a Team plan afterwards does not add the permission. Disconnect Tactiq in Claude, reconnect, and confirm the meeting details permission is selected on the Tactiq approval screen. An active Team seat is required for it to appear.
How do I revoke an AI assistant's access to my Tactiq meetings?
Revoke it from Tactiq rather than from the assistant. Open Settings then MCP in the Tactiq app, find the connection under active connections, and revoke it. Tactiq states this immediately invalidates that assistant's access. Revoking is also the documented fix when you need the authorisation screen to appear again with a different permission set.
Sources
- Tactiq MCP connection guide, for the plan table, the server URL, the four permissions and their descriptions, the seven documented capabilities, the revocation steps and the three common errors (retrieved 2026-08-19; Anthropic's published
documentationURL with the-tactiq-mcp-serverslug 301-redirects here) · retrieved 2026-08-19 - Tactiq Claude setup guide, for the directory and custom-connector steps, the Claude Team and Enterprise owner requirement, the shared-meetings note, the generation-delay note and the permissions-do-not-upgrade FAQ answer (retrieved 2026-08-19) · retrieved 2026-08-19
- Tactiq LLM privacy article, for Microsoft Azure OpenAI as the model provider, the no-training and disabled-content-logging statements, and the 30-day archive deletion (retrieved 2026-08-19) · retrieved 2026-08-19
- Tactiq transcript deletion article, for the archive-then-delete flow, the 30-day window, the irreversibility of permanent deletion and account-deletion behaviour (retrieved 2026-08-19) · retrieved 2026-08-19
- Tactiq transcription notification article, for the opt-in privacy setting and the one-time chat message it posts (retrieved 2026-08-19) · retrieved 2026-08-19
- Tactiq roles and permissions article, for the three Space permission levels (retrieved 2026-08-19) · retrieved 2026-08-19
- Tactiq pricing page, for the plan prices and inclusions, the Business-plan Tactiq MCP (beta) and Tactiq Claude Connector (beta) rows, the AI credit definitions and the OpenAI enterprise API answer (retrieved 2026-08-19) · retrieved 2026-08-19
- Tactiq Privacy Center, for the OpenAI enterprise API statement that its LLM article contradicts (retrieved 2026-08-19) · retrieved 2026-08-19
- Tactiq
llms.txt, for the bot-free browser capture, the no-audio-stored statement and the plan summary (retrieved 2026-08-19).tactiq.io/robots.txtexplicitly allows ClaudeBot, GPTBot, anthropic-ai, PerplexityBot, Google-Extended and CCBot by name. · retrieved 2026-08-19 - Live auth posture check against
https://mcp.tactiq.io/— an anonymousinitializereturned HTTP 401 with aWWW-Authenticate: Bearerheader namingmissing_token, aresource_metadatapointer and all four scopes. The RFC 9728 descriptor athttps://mcp.tactiq.io/.well-known/oauth-protected-resourcereturned HTTP 200 withresourceandauthorization_serversboth naminghttps://mcp.tactiq.ioand fourscopes_supportedentries; the trailing-slash form returned an identical body and the path-insert form 404'd. RFC 8414 metadata at/.well-known/oauth-authorization-serverand an identical document at/.well-known/openid-configurationreturned HTTP 200 declaring PKCES256, dynamic client registration andservice_documentationathttps://tactiq.io/help/mcp(2026-08-19) · retrieved 2026-08-19 - Anthropic connector directory entry for Tactiq, for the nine tool names, the partner tier, the endpoint, the empty
prompt_nameslist, theworks_withclient list and the absence of anypermissionsfield (snapshot 2026-08-16) · retrieved 2026-08-16
Use in Agentman
Connect once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.
Open in Agentman StudioServer Info
- Category
- Productivity
- Developer
- Tactiq
- Tools
- 9
- Domain
- mcp.tactiq.io
Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.