Asana MCP Server

The Asana MCP Server is Asana's official Model Context Protocol server. It connects MCP-capable AI clients such as Claude, Claude Code, Cursor, ChatGPT, or Codex to the Asana Work Graph, enabling users to read, create, and edit tasks, projects, portfolios, and goals directly through natural language. The server is available exclusively as a hosted remote service at https://mcp.asana.com/v2/mcp. According to the provider, this is a V2 server using Streamable HTTP transport (MCP specification v2). No public GitHub repository is provided for this server.

What is the Asana MCP Server

The Asana MCP Server gives AI agents structured access to the Asana Work Graph — the data model that links tasks, projects, portfolios, goals, sections, comments, attachments, and user profiles. According to the provider, the server supports natural language queries such as: "Show me all my open tasks due this Friday", "Create a new task in the Marketing project", "List all sections in the Product Launch project", or "Give me a status overview of the Q2 Planning project". The server translates these requests into structured MCP tool calls executed directly against the Asana API.

Prerequisites

Using the Asana MCP Server requires an active Asana account. According to the provider, the server supports all MCP clients that implement Streamable HTTP transport. OAuth is used for authentication; on first connection, the client opens a browser login dialog for Asana authorization. In Enterprise+ and Legacy Enterprise workspaces, Asana administrators can restrict or block access via app management. Workspaces on lower plans can contact support to block beta access. The advanced search tool (search_tasks) is, according to the provider, restricted to Premium accounts; get_tasks is available as an alternative for Standard accounts.

Features: Tasks, Projects, and the Work Graph

The V2 server provides three categories of tools according to the provider:

Read tools return data from the workspace without making any changes: search_objects for universal search across tasks, projects, portfolios, goals, teams, users, and AI Teammates; get_task and get_tasks for task details and filtered task lists; get_my_tasks as a shortcut for the user's own task list; get_project and get_projects for project details and lists; get_portfolio and get_portfolios for portfolio data; get_items_for_portfolio for portfolio contents; get_status_overview for aggregated status reports; get_attachments for file attachments; and tools for users, teams, sections, comments, and subtasks.

Write tools create, update, or delete data: create_task, update_task, create_project, update_project, and tools for comments, attachments, subtasks, and section memberships.

Interactive tools display a confirmation UI before committing a change, according to the provider. This feature is currently available in Claude and ChatGPT.

Setup

The Asana MCP Server is available at https://mcp.asana.com/v2/mcp. For Claude Code, the setup command is:

claude mcp add --transport http asana https://mcp.asana.com/v2/mcp

For other MCP clients, the server is added in the client-specific configuration file (e.g. .mcp.json or settings.json) under the URL https://mcp.asana.com/v2/mcp as a remote server with HTTP transport. According to the provider, Claude, ChatGPT, and many coding-focused clients support direct connection via OAuth. After connecting, a browser login dialog for Asana authorization appears on first use. The full list of compatible clients and client-specific instructions is available in the official documentation.

Security and Data Access

According to the provider, the Asana MCP Server does not use a separate permission scope model. When a user authorizes an MCP client, the token automatically requests full access to all available tools. Within that token, however, the server can only access tasks, projects, and data for which the authenticated user already has read permission in Asana. Write tools are subject to the same permission boundaries. According to the provider, the server does not forward Asana data directly to AI models; the connected AI client decides which tool results are passed to its model. Enterprise+ and Legacy Enterprise administrators can control MCP access per client through app management.

FAQ

Is the Asana MCP Server free? Access to the MCP Server is available with an existing Asana account. Some advanced search features (e.g. search_tasks) are, according to the provider, restricted to Premium accounts. Current plan and pricing details are available at asana.com/pricing.

Can I run the server locally? No. According to the provider, the Asana MCP Server is available exclusively as a hosted remote service. There is no public repository or local installation package.

Which MCP clients are supported? All MCP clients with Streamable HTTP transport support. According to the provider, Claude, Claude Code, ChatGPT, Cursor, Codex, VS Code, and other clients are compatible. Clients without OAuth support can build their own integration using a client ID and client secret following the official integration guide.

What is the Work Graph? The Asana Work Graph is the data model that relates and contextually links tasks, projects, portfolios, goals, teams, and users. The MCP Server uses this model to give AI agents structured access to work context.

Active Asana account and an MCP-capable AI client supporting Streamable HTTP transport. A Premium account is required for search_tasks.

For Claude Code: claude mcp add --transport http asana https://mcp.asana.com/v2/mcp. Configure other clients via .mcp.json or client settings with the URL https://mcp.asana.com/v2/mcp (HTTP transport). Complete the OAuth login on first connection.

claude mcp add --transport http asana https://mcp.asana.com/v2/mcp

OAuth via Asana login. On first connection, the MCP client opens a browser dialog for authorization.

Full access to all MCP tools after authorization; the server is limited to data the authenticated user has permission to access in Asana. No scope model.

The server returns tool results; the connected MCP client decides which data is forwarded to the AI model.

Since all tools are accessible after authorization, the MCP client should only be used in trusted environments. Write tools can modify tasks and projects without additional confirmation (except for interactive tools in Claude/ChatGPT).