Integration Model Context Protocol OAuth 2.1

Wizideo MCP Server

Connect Claude, Claude Code, Cursor and any MCP-compatible agent directly to your Wizideo context — meetings, transcripts, summaries, topics, boards, tasks and insights. One secure endpoint, one-click OAuth sign-in. No API keys to copy.

Add to your agent How it works
Endpoint https://mcp.wizideo.ai/mcp

What is the Wizideo MCP server?

The Model Context Protocol (MCP) is an open standard that lets AI assistants talk to external tools and data through one consistent interface. Wizideo exposes its meeting intelligence as an MCP server, so your AI agent can read transcripts, search across every meeting, draft follow-ups and create tasks — without copy-pasting anything.

It runs over Streamable HTTP at a single endpoint and authenticates with OAuth 2.1 — your client opens Wizideo's consent screen, you pick the team and approve exactly which scopes to grant. No API keys to create or paste. Every write is recorded in an audit log.

  • ✓ Works with any MCP client — Claude.ai, Claude Code, Cursor, Claude Desktop.
  • ✓ One-click OAuth — approve read-only, task-write or integration-push on a consent screen.
  • ✓ Semantic search over your entire meeting history.
  • ✓ Audit trail on every write, rate-limited and team-isolated.
Claude
Claude Code
Cursor
Your agent
MCP mcp.wizideo.ai
Transcripts
Summaries
Boards & Tasks
Topics & Insights

What your agent can do

Each tool maps to a scope you approve at sign-in. Grant only what you need — the server rejects any call outside the granted scopes with 403 insufficient_scope.

Read

mcp:read

Browse and search everything in your Wizideo workspace.

  • list_projectsList the user's meeting projects (optional folder filter).
  • get_projectFull detail of a project by UUID.
  • search_projectsSearch projects by name.
  • get_transcriptionFull meeting transcript — text, segments or SRT, with time-range slices.
  • search_transcriptionsSemantic search across every transcript (Qdrant + embeddings).
  • list_summariesAI summaries generated for a project.
  • get_summaryFull content of a single summary.
  • list_topicsAuto-detected topics for the team.
  • get_topicDetail of a topic.
  • get_boardAction board (actions + commitments) of a meeting.
  • list_tasksTeam tasks, filterable by project, status, assignee or kind.
  • get_taskDetail of a single task.
  • list_insightsAnalyzer executions for a project.
  • get_insightFull result of an analyzer execution.

Write tasks

mcp:tasks:write

Let the agent create and update tasks on your meeting boards.

  • create_taskCreate a manual task on a meeting board.
  • update_taskPartial-update an existing task.
  • mark_task_readyTransition a task from draft to ready.

Push to integrations

mcp:integrations:push
  • push_task_to_integrationSend tasks to Trello, Linear, Jira… — idempotent, by slug or id.

Plus stable resources (wizideo://project/{uuid}, …/transcript, …/board) and ready-made prompts: wizideo.recap, wizideo.followup, wizideo.action_items.

Connect in 3 steps

1

Add the connector

Point your MCP client at https://mcp.wizideo.ai/mcp — one command in Claude Code, one line of JSON in Cursor, or "Add custom connector" in Claude.ai. Configs below.

2

Sign in & approve

Your client opens Wizideo's OAuth consent screen in the browser. Sign in, choose the team to expose and approve the scopes the client asked for. Click Allow.

3

Start working

That's it — the client holds a scoped access token and the team is signed into it (no headers to set). Ask your agent anything about your meetings.

Connect your client

Same endpoint everywhere: https://mcp.wizideo.ai/mcp. Authentication is handled by OAuth — your client opens the consent screen the first time you use it. No headers, no keys.

Add the server with a single CLI command in your project:

terminal 
claude mcp add --transport http wizideo https://mcp.wizideo.ai/mcp

Then run /mcp inside Claude Code and choose Authenticate — it opens the browser for OAuth sign-in. Add --scope user to share the server across all your projects. Verify with claude mcp list.

Add to ~/.cursor/mcp.json (global) or .cursor/mcp.json (per-project):

.cursor/mcp.json 
{
  "mcpServers": {
    "wizideo": {
      "url": "https://mcp.wizideo.ai/mcp"
    }
  }
}

Open Cursor Settings → MCP, click Login next to wizideo to complete OAuth, then confirm it shows a green dot with its tools.

Add a Custom Connector in Claude.ai (Pro, Max, Team or Enterprise):

  1. Go to Settings → Connectors → Add custom connector.
  2. Name it Wizideo.
  3. URL: https://mcp.wizideo.ai/mcp
  4. Save, then click Connect — Wizideo's consent screen opens. Pick your team, approve the scopes and click Allow.
  5. Open any chat — Wizideo appears in the Tools menu.

Claude Desktop reaches HTTP servers through mcp-remote, which handles the OAuth flow for you. Edit claude_desktop_config.json:

claude_desktop_config.json 
{
  "mcpServers": {
    "wizideo": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.wizideo.ai/mcp"]
    }
  }
}

Restart Claude Desktop — a browser tab opens for OAuth sign-in the first time. After approving, look for the tools icon in the composer.

Try it

Once connected, talk to your agent in plain language. It picks the right tools.

Recall

"List my last 5 meetings and summarize what each one decided."

Search

"Find every meeting where we discussed the onboarding revamp."

Draft

"Draft a follow-up email for my sync with Acme this morning."

Act

"Create tasks for every action item from yesterday's standup and push them to Linear."

Rate limits

Sliding-window buckets per user & team. Exceeding one returns 429 with Retry-After.

  • Read600 / min
  • Search60 / min
  • Write30 / min

Built-in safeguards

  • ✓ OAuth 2.1 with PKCE — short-lived tokens, revocable instantly.
  • ✓ Team isolation — the team is signed into the token.
  • ✓ Transcripts capped at ~30k tokens with range slicing.
  • ✓ Every write logged (SHA-256 of args, no plaintext).

Privacy & data

MCP calls only ever send your AI client exactly what a tool returns — no raw embeddings, no internal metadata. Write tools require their own scope and leave an audit trail. We don't use your meeting data to train third-party models.

Privacy Policy →

Plug Wizideo into your AI agent

Add the endpoint to your client, approve access with one click, and let your agent work with your meetings.

Add to your agent Create a free account

Need help? Contact us at support@wizideo.ai or visit our contact page.