mcpField
C
claws-mail-mcp-plugin

claws-mail-mcp-plugin

Local
github.comCommunication0Updated 1w ago

About claws-mail-mcp-plugin

A native Claws Mail plugin that exposes the running mail client to local MCP-capable AI tools. It supports message and folder operations, with a read-only mode for restricting changes.

Key features

  1. 01Read messages and folders
  2. 02Search, compose, and send mail
  3. 03Reply to and forward messages
  4. 04Manage folders and message flags
  5. 05Local HTTP endpoint with token protection
  6. 06Configurable read-only mode

Use cases

  1. 01Let an AI tool search and read mail
  2. 02Compose or send mail through configured accounts
  3. 03Organize folders and flags in the mail client

MCP support

The plugin runs inside Claws Mail and provides a loopback-only MCP HTTP endpoint protected by a bearer token. MCP clients can use its tools to read, search, and manage mail.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

No install method listed yet

We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.

Repository

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

claws-mail-mcp-plugin

CI Claws Mail License: GPL v3+

Native Claws Mail plugin that exposes the running mail client as an MCP server.

It runs inside Claws Mail and serves a local, loopback-only HTTP endpoint (JSON-RPC over MCP, bearer-token protected). Any MCP-capable AI tool can then read folders and messages, search, compose, reply, forward, manage folders and flags, and send or receive mail through your already-configured accounts. Endpoint, port, token and read-only mode are set in mcp_pluginrc inside the Claws Mail config directory; the plugin writes a ready-to-paste client config to mcp_client.json in the same place on every start.


Installing the plugin

Prebuilt binaries are attached to each release, one per OS and supported Claws Mail version (e.g. mcp-linux-claws4.3.1.so, mcp-windows-claws4.3.0.dll). Pick the file matching your Claws Mail version and OS, then:

  1. Save it in a folder of your choice, for example ~/.claws-mail/plugins or %APPDATA%\Claws-mail\plugins on Windows, as mcp.so or mcp.dll. Any file name works except one ending in _plugin.so or _plugin.dll: Claws Mail does not load such a file on startup and drops it from the plugin list.
  2. In Claws Mail, open Configuration > Plugins, click Load... and select the file.

Claws Mail loads the plugin again on every start.

To build from source instead, see Building.


Connecting an AI tool

Once loaded, the plugin writes the real endpoint URL and bearer token to mcp_client.json in the Claws Mail config directory. Use those two values with whichever method below matches your tool. The server is loopback-only (127.0.0.1), so it is reachable only from AI tools running on the same machine as Claws Mail.

Claude Code
claude mcp add --transport http claws-mail http://127.0.0.1:8765/mcp \
  --header "Authorization: Bearer <TOKEN>"
Claude Desktop

Settings > Developer > Edit Config, then merge in packaging/claude/mcp.json (fill in <TOKEN>). Do not use the "Add custom connector" UI for this server: that flow connects from Anthropic's servers, not from your machine, so it cannot reach a loopback address.

VS Code / GitHub Copilot Chat
code --add-mcp "{\"name\":\"claws-mail\",\"type\":\"http\",\"url\":\"http://127.0.0.1:8765/mcp\",\"headers\":{\"Authorization\":\"Bearer <TOKEN>\"}}"

Or copy packaging/vscode/mcp.json to .vscode/mcp.json; it prompts for the token instead of storing it in the file.

Codex CLI
export CLAWS_MCP_TOKEN=<TOKEN>
codex mcp add claws-mail --url http://127.0.0.1:8765/mcp --bearer-token-env-var CLAWS_MCP_TOKEN

This matches packaging/codex/config.toml, which Codex reads from ~/.codex/config.toml.

MCP registry

The plugin is listed in the MCP registry as io.github.it-atelier-gn/claws-mail-mcp-plugin. The listing carries no connection details; install the plugin and connect your tool as described above.


Building

Requires Meson, Ninja, a C compiler, and the Claws Mail development files (GTK3, GLib, GnuTLS, libgcrypt; on Linux also enchant).

Linux

Install dependencies (Debian/Ubuntu example):

sudo apt-get install build-essential meson ninja-build pkg-config \
  libgtk-3-dev libglib2.0-dev libgnutls28-dev libgcrypt20-dev libenchant-2-dev

If a claws-mail pkg-config file is not already on your system, build one with:

PREFIX="$PWD/.claws-sdk" ci/build-claws-sdk.sh
export PKG_CONFIG_PATH="$PWD/.claws-sdk/lib/pkgconfig"

Then configure and build:

meson setup build
meson compile -C build
meson test -C build

The plugin is built at build/mcp.so.

Windows (MSYS2)

Install the MINGW64 toolchain and dependencies:

pacman -S mingw-w64-x86_64-gcc mingw-w64-x86_64-meson mingw-w64-x86_64-ninja \
  mingw-w64-x86_64-pkgconf mingw-w64-x86_64-gtk3 mingw-w64-x86_64-glib2 \
  mingw-w64-x86_64-gnutls mingw-w64-x86_64-libgcrypt

Build the Claws Mail SDK and set PKG_CONFIG_PATH as above, then run the same meson setup / meson compile / meson test commands. The plugin is built at build/mcp.dll.

Plugin version

The version compiled into the binary (returned by plugin_version and the status tool) comes from -Dplugin_version, e.g.:

meson setup build -Dplugin_version=1.2.3

CI sets this from the pushed git tag on release builds. Without it, the build falls back to git describe, then to 0.0.0-dev.

Installing
meson install -C build

This installs the plugin into the Claws Mail plugin directory reported by pkg-config --variable=plugindir claws-mail. Load it from there once under Configuration > Plugins > Load...


Supported Claws Mail versions

Each release contains a Linux and a Windows build for every Claws Mail version listed in ci/claws-versions.json.


License

GPL-3.0-or-later, see LICENSE.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.