Connect AI agents to CalmBoard with MCP

CalmBoard paid plans include an MCP server so AI agents and MCP clients can inspect real project data without screen scraping or manual exports. The server exposes project, board, task, and AI insight tools through the same permission model used by CalmBoard integrations.

MCP AI Agents Integrations OAuth

What the CalmBoard MCP server does

The MCP server gives an AI agent a structured way to ask for CalmBoard context. Instead of giving an agent a raw database export, you connect it to https://calmboard.ai/mcp and let it call named tools such as calmboard.list_projects, calmboard.list_boards, calmboard.list_tasks, and calmboard.list_insights.

Access is tied to the CalmBoard user who creates the API key or authorizes the OAuth flow. The agent can only retrieve data that user can already access, and the feature is available only on paid plans.

Setup checklist

Open the private Integrations & APIs page in CalmBoard.
Create an API key, or copy your OAuth client ID and save your allowed redirect URI.
Point your MCP client to https://calmboard.ai/mcp.
Authenticate with Authorization: Bearer cbk_your_api_key or an OAuth-issued cbo_ bearer token.

Example 1: inspect tasks from an MCP client

After the client initializes the MCP session, call the task tool with a project and board ID:

{
  "jsonrpc": "2.0",
  "id": 2,
  "method": "tools/call",
  "params": {
    "name": "calmboard.list_tasks",
    "arguments": {
      "project_id": "prj_123",
      "board_id": "brd_123",
      "limit": 20
    }
  }
}

This is useful for questions like "which tasks are blocked this week?" or "what work is still in review?" because the agent receives structured task data instead of guessing from a static screenshot.

Example 2: ask an agent to summarize delivery risk

Give your agent permission to call calmboard.list_insights and calmboard.list_tasks. A typical instruction can be:

Review the current board tasks and recent CalmBoard AI insights. Summarize the top delivery risks in plain language, list the tasks that need attention, and avoid ranking people or comparing team members.

The agent can then combine CalmBoard's existing AI insight cards with live board state, while still respecting board access and plan limits.

OAuth discovery for AI agents

MCP-capable clients can discover CalmBoard OAuth metadata from:

GET https://calmboard.ai/.well-known/oauth-protected-resource/mcp
GET https://calmboard.ai/.well-known/oauth-authorization-server
POST https://calmboard.ai/api/oauth/register

Use OAuth when you want a user-consented connection. Use API keys for controlled service integrations where a CalmBoard user intentionally creates and manages the key.

Back to blog Open Integrations & APIs