What it does
Flutter MCP Toolkit is a Dart MCP server paired with a Flutter package. It lets an AI assistant inspect a running Flutter app, take semantic snapshots, search the widget tree, tap widgets, type into forms, hot-reload and read logs. The app can also register its own MCP tools and resources at runtime through the mcp_toolkit package.
It suits a developer who wants an AI assistant to check a Flutter app's actual state while debugging, instead of guessing from source code. The agent can observe the app, act on it and verify the result in one loop. For scripts and CI, the same operations are available from a command-line interface (flutter-mcp-toolkit or fmtk), and repeatable scenarios can be written as Dart code with flutter_mcp_harness.
Setup depends on the client. The toolkit can be initialised for any agent, installed as a Claude Code plugin or a Codex or Cursor plugin, or added as skills only. It is also listed in MCP registries, where the server is served as an OCI image from GHCR. Named clients include Codex, Zed, Cursor, Claude Code and Cline.
Limitations and requirements
Designed for running Flutter debug apps and games. The server binary is shipped as GitHub Release artifacts and is not a pub.dev package. Apps that expose their own custom MCP tools need the mcp_toolkit Flutter package added to the app.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 14 Nov 2026 · its page on the source
Flutter MCP Toolkit
Inspect and drive Flutter debug apps/games: semantic snapshots, search, and custom client tools
v7.1.1As shown on registry.modelcontextprotocol.io when we last read it