projio

Project knowledge orchestrator and MCP server for research repositories.

Turns a research repository into a queryable knowledge environment for humans and AI agents. Provides structured, machine-accessible knowledge layers by integrating code, papers, notes, pipelines, figures, and documentation through a unified MCP server interface.

Install

pip install projio                # core orchestrator + MCP server
pip install "projio[all]"         # all ecosystem packages

For development (pixi — primary):

git clone https://github.com/arashshahidi1997/projio.git
cd projio
pixi install -e dev            # install projio + all subpackages + dev tools
pixi run -e dev test           # run the test suite
pixi run docs                  # build the docs site

Conda is also supported. See CLAUDE.md for the make dev / make test path and the code.runner: conda project-level fallback.

Quick start

# Scaffold .projio/ workspace
projio init .

# Scaffold a study or tool project
projio init . --kind study
projio init . --kind tool

# Show project status
projio status -C .

# Start MCP server (stdio)
projio mcp -C .

# Site operations
projio site build
projio site serve

Ecosystem

Projio orchestrates six specialized packages:

Install individual extras: pip install "projio[biblio]", pip install "projio[pipeio]", etc.

MCP tools

The MCP server exposes 70+ tools across all subsystems:

See the MCP tools reference for the full list.

Configuration

Project-local defaults live in .projio/config.yml. Optional user defaults in ~/.config/projio/config.yml are merged first, with project config taking precedence.

# Scaffold user config
projio config init-user

# Show merged config
projio config -C . show

Claude Code / MCP config

{
  "mcpServers": {
    "projio": {
      "command": "/path/to/python",
      "args": ["-m", "projio.mcp.server"],
      "env": { "PROJIO_ROOT": "/path/to/your/project" }
    }
  }
}

Documentation

Full docs at arashshahidi1997.github.io/projio.

License

MIT