Anytype Mcp
Connect your agent to an encrypted Anytype workspace so specs, research, and task notes stay in sync while you code.
Overview
Anytype MCP is a MCP server for the Build phase that connects AI assistants to your encrypted Anytype wiki through the official Anytype API.
What is this MCP server?
- Official Anytype API MCP server (@anyproto/anytype-mcp v1.0.7)
- Encrypted, local-first wiki access for AI assistants over stdio
- Auth via OPENAPI_MCP_HEADERS JSON (Bearer token plus Anytype-Version header)
- Bridges agent workflows to structured objects and pages—not a generic markdown folder sync
- Package version 1.0.7 (@anyproto/anytype-mcp on npm)
- Transport: stdio
- Required secret header bundle: OPENAPI_MCP_HEADERS
Community signal: 437 GitHub stars.
What problem does it solve?
Your product knowledge lives in Anytype while your agent only sees the repo, so planning context never rides along with implementation tasks.
Who is it for?
Builders who already run Anytype as an encrypted PKM or team wiki and want Claude Code or Cursor to pull tasks and references from that graph.
Skip if: Teams on Notion-only or plain-git-docs workflows with no Anytype API setup, or users who need full UI parity inside the agent.
What do I get? / Deliverables
Agents can query and update Anytype content in chat so specs and notes stay aligned with what you ship.
- Agent-driven reads and writes against Anytype objects via MCP tools
- Unified dev session where implementation tasks can cite live wiki state
Recommended MCP Servers
Journey fit
Knowledge bases and living documentation are central during build when scope, architecture, and decisions accumulate faster than repo READMEs can carry. The docs subphase is where wikis, specs, and decision logs live—Anytype is explicitly positioned as a collaborative encrypted wiki.
How it compares
Official wiki API MCP server, not a markdown repo indexer or generic note-taking skill.
Common Questions / FAQ
Who is Anytype MCP for?
Solo builders and small teams who use Anytype as their encrypted knowledge base and want MCP-capable agents to work against live workspace data.
When should I use Anytype MCP?
Use it during build and ongoing operate when specs, roadmaps, or research in Anytype should inform code changes without manual export.
How do I add Anytype MCP to my agent?
Install @anyproto/anytype-mcp from npm, configure stdio transport, and set OPENAPI_MCP_HEADERS to a JSON string with Authorization Bearer and Anytype-Version per Anytype API docs.