MCP Hub

Your AI operates Orkestra. Paste one URL, done.

Connect claude.ai or ChatGPT with OAuth 2.1: no tokens to copy, no files to edit. Once authorized, your AI operates 99% of Orkestra in natural language.

Connect your AI See MCP docs

What is MCP?

Model Context Protocol (MCP) is an open standard created by Anthropic that lets any AI assistant connect to external services through a unified protocol. Instead of re-inventing integrations for every AI, you just expose an MCP server — and all compatible agents can talk to it.

Orkestra was built with MCP in mind from day one. Instead of flooding your AI with hundreds of tools, we expose 29 domain tools (tasks, projects, wiki, sprints, finance, stock and more). Each takes an action parameter that unlocks around 296 actions, and action: "describe" returns the full schema on demand. The tool list stays small — up to 4× less client context than one tool per action.

Every write operation goes through preview/confirm, so agents never execute destructive changes without validation. On top of that, every user can fill in a personal AI context (role, expertise, preferred tone, timezone) in their profile. The MCP server exposes it as the resource orkestra://user/context and through the orkestra_user → get_profile action, so your agents know who you are from the first message — no need to repeat yourself in every conversation.

29domain tools
~296actions
99%of the backend
2transports (STDIO + HTTP)
OAuth · no token

claude.ai and ChatGPT

The fastest path: paste one URL and OAuth 2.1 does the rest. No tokens to copy, no config files to edit.

Connector URL text 
https://mcp.orkestra.team/mcp
  1. In claude.ai: Settings → Connectors → Add custom connector. In ChatGPT: connectors in developer mode.
  2. Paste the URL https://mcp.orkestra.team/mcp.
  3. Log in to Orkestra and accept the consent screen.
  4. Done. Future reconnections are silent — once approved, you're never asked to consent again.

Prefer a guided walkthrough from the app? Open Connect your AI.

OAuth · terminal

Claude Code

From the terminal, a single command. OAuth authorization opens in your browser.

terminal bash 
claude mcp add --transport http orkestra https://mcp.orkestra.team/mcp
  1. Run the command above.
  2. Your browser opens: log in to Orkestra and authorize.
  3. Return to the terminal — Orkestra is connected.
Token · local and custom

Local and custom clients

For local STDIO MCP servers (Claude Desktop) or any custom HTTP client (N8N, scripts, your own agent) that sends Authorization: Bearer ork_..., you still use tokens. Each token carries an access level: full or read-only (read-only tokens can call only read tools).

claude_desktop_config.json (STDIO) json 
{
  "mcpServers": {
    "orkestra": {
      "command": "npx",
      "args": ["-y", "@orkestra/mcp-server"],
      "env": {
        "MCP_TOKEN": "ork_your_token_here"
      }
    }
  }
}
HTTP client (N8N, scripts) json 
{
  "mcp_servers": {
    "orkestra": {
      "transport": "http",
      "url": "https://mcp.orkestra.team/mcp",
      "headers": {
        "Authorization": "Bearer ork_your_token_here"
      }
    }
  }
}
  1. Go to Settings → MCP Tokens and create a token (choose full or read-only access).
  2. Copy the value (shown only once).
  3. Paste it into the MCP_TOKEN env var (STDIO) or the Authorization: Bearer header (HTTP), replacing ork_your_token_here.
  4. Restart the client. Done.

Context, prompts and real-time

Resources that load themselves

Compatible clients auto-load context: orkestra://user/myday (today's tasks and activities), orkestra://user/organizations (org and project map), orkestra://user/context (your AI context) and, at org level, world-model (DRIs, decisions, gaps) and structure (areas and org chart).

5 ready-made prompts

The server ships commands that compatible clients surface directly: Weekly summary, Plan sprint, Backlog triage, My day and Project status report.

Changes in real time

When the AI writes, the web app updates instantly (Socket.io) and webhooks, Slack and automation rules fire just as if the change came from the app.

Security and control

Automatic preview / confirm

Every write returns a preview with an actionId; the agent confirms with the orkestra_confirm_action tool. In STDIO clients with elicitation (Claude Desktop, Claude Code) the confirmation appears as a native dialog. Limit: 30 confirmations/min per user.

Access level per token

Each token is issued with full or read-only access. A read-only token can't execute any write action, even if the agent tries.

Hashed tokens

We only store the SHA-256 hash of the token. The raw value is shown once at creation.

Instant revocation

Revoke a token from the UI and active connections close — revoking also disconnects OAuth clients. No exposure windows.

Per-user scope

Each connection acts on behalf of the user who authorized it — with all their permissions and visibility, not one more.

Start operating with AI today

MCP is included for everyone, at no extra cost. No tricks, no fine print.

Create free account See all tools