mcp-executor

# mcp-executor `mcp-executor` is a lightweight FastMCP driver that discovers every MCP tool exposed by your configured backends, prints Python-like signatures, and lets you invoke them either from the command line or through a minimal FastMCP server. ## Why use it? - 📋 **Discoverability**: Generate readable signatures for every tool so you know which arguments to pass. - ⚡ **Fast iteration**: Jump between running the HTTP/stdio server and issuing one-off tool calls without reconfiguring clients. - 🧰 **Batteries included**: Ships with a `uv`-first workflow for building, testing, and publishing to PyPI. ## Installation ```bash pip install mcp-executor ``` or, if you prefer the `uv` toolchain: ```bash uv tool install mcp-executor ``` ## Configuration Point the CLI at an MCP client definition. Copy `mcp.json.example` into your workspace, adjust the upstream server definitions, and pass the path with `-c/--config`. ```bash cp mcp.json.example ~/.config/mcp-executor/mcp.json ``` ## Usage List available tools and inspect their signatures: ```bash mcp-executor list -c ~/.config/mcp-executor/mcp.json ``` Call a tool directly: ```bash mcp-executor call -c ~/.config/mcp-executor/mcp.json weather --arg city="Lisbon" ``` Run the FastMCP server (HTTP by default): ```bash mcp-executor serve -c ~/.config/mcp-executor/mcp.json --transport http --host 0.0.0.0 --port 23456 ``` Every command shares the `--config` option so you can point at different MCP client definitions per invocation. ## Local development ```bash uv venv --seed uv sync uv run mcp-executor list -c mcp.json.example ``` The repository still ships `main.py` so you can run `./main.py list ...` directly with `uv run` if you prefer scripting locally. ## Releasing with `uv` 1. Bump `version` inside `pyproject.toml`. 2. Build and verify the artifacts: ```bash uv build uv run python -m mcp_executor.cli --help ``` 3. Publish to PyPI with an API token (create one under your PyPI accou…