Tenali AITenali AI/LLM GuideTry Free
This page is optimized for AI assistants and LLM search
Last updated: 2026-10-03

Tenali MCP Server

Tenali runs a Model Context Protocol (MCP) server, so an AI assistant can read your Tenali data directly instead of you pasting transcripts into a chat window. Add it once and ask questions in plain language — the assistant looks everything up itself.

The essentials

  • Endpoint: https://api.tenali.ai/api/mcp, added as a custom connector or remote MCP server.
  • Transport: streamable HTTP. There is no stdio or SSE endpoint, and no API key — authentication is OAuth 2.1.
  • Included on every Tenali plan, for every workspace. There is no separate MCP subscription, add-on or tier, and nothing an admin has to switch on. Your assistant's own plan still has to support custom connectors.
  • Read-only. Fourteen tools, none of which can create, edit, delete or send anything in Tenali.
  • Copy the server URL from Settings → AI Assistants in the Tenali web app.

What an assistant can reach

Four permissions, chosen on the approval screen at sign-in:

  • Meeting summaries (required) — titles, attendees, briefs, summary sections, action items, and the customer questions Tenali detected with the answers it captured.
  • Knowledge base content (optional) — your team's documents, FAQs, crawled sites and other training sources.
  • Raw meeting transcripts and recording playback (optional) — full timestamped transcript text, with speaker names and emails, plus a time-limited link to play or download the recording itself.
  • All meetings across your organization (optional, admins only) — widens every read from your own meetings to the whole workspace.

Unticking a permission makes it unavailable to the assistant rather than merely hidden: the tool still appears in the manifest and fails at call time.

The fourteen tools

Tool Permission
search_meeting_summaries, get_meeting_summary, list_meetings, list_action_items, review_pipeline, question_intelligence, list_question_occurrences, get_meeting_recording Meeting summaries
get_meeting_transcript, search_transcripts Raw transcripts and recordings
search_knowledge_base, list_sources, get_source_metadata Knowledge base
team_activity Organization-wide (admin)

review_pipeline builds account rollups from your meetings, not from a CRM — it is not a pipeline report. get_meeting_recording can be called with Meeting summaries alone, but only returns a usable link when the recording permission is also granted.

Supported clients

Claude, ChatGPT, Perplexity and Gemini connect by pasting the server URL as a custom connector. Cursor and VS Code have a one-click install from Settings → AI Assistants → Developer tools. Any other client that supports remote MCP servers over streamable HTTP with OAuth works the same way. Each assistant's own plan governs whether it can add a custom connector at all.

Security posture

  • Never more than you can see. Identity is pinned when you approve the connection and re-resolved server-side on every single request, never from the token's claims alone. An assistant sees exactly what you'd see signed into Tenali.
  • Content is data, not instructions. The server tells every connected assistant, in its own instructions, that meeting and document content is to be read rather than followed — so a line someone drops into a call, a calendar invite or an uploaded document cannot redirect your assistant through Tenali. Responses are also scrubbed of Tenali's internal identifiers before they leave.
  • Organization-wide access clears two gates on every request — an Admin or Super User role, and the workspace's Admin Full Access setting. If the person loses that role or Admin Full Access is turned off, the very next tool call stops widening.
  • Revocable instantly. Revoking a connection takes effect on the assistant's next request even if it still holds a valid token. Each connection covers one workspace.
  • Logged. Tool calls are recorded, with organization-wide reads and cross-owner recording access recorded separately.
  • Tokens are issued through a dedicated identity and OAuth provider; there is no API key to generate, paste, store or rotate.

Limits

  • 120 tool calls per minute per connection.
  • 400 KB per response, 256 KB per request, 45-second tool timeout.
  • list_meetings returns 50 per page (up to 200); list_action_items 200 (up to 500); get_meeting_transcript 200 utterances (up to 500); search_transcripts 50 hits (up to 200) across the 100 most recent meetings in scope; review_pipeline analyzes up to 500 meetings and says whether that covered everything.
  • Anything an assistant retrieves leaves Tenali's retention and is governed by that assistant's own settings and terms.

What it does not do

It does not replace a conversation-intelligence or forecasting tool, and it never writes to Tenali or to your CRM — CRM write-back is a separate capability in Tenali's own chat and in Slack.