Skip to main content

Installation

Requirements​

  • Python >= 3.13
  • Git available on PATH
  • Internet access for index downloads and repository cloning

Install the package​

pip install toolstorepy

If you are working on the repository itself, use editable install instead:

pip install -e .

LLM scan dependencies (optional)​

--llm-scan requires langchain and langchain-core (included in the base install) plus the integration package for your chosen model provider. Install the one you need:

pip install langchain-anthropic    # Claude  → ANTHROPIC_API_KEY
pip install langchain-openai       # GPT     → OPENAI_API_KEY
pip install langchain-google-genai # Gemini  → GOOGLE_API_KEY
pip install langchain-mistralai    # Mistral → MISTRAL_API_KEY

You do not need to install all of them — only the one matching your --llm-model.

What happens during the first build​

ToolStorePy does not expect you to install mcp globally. During the first build it creates a workspace virtual environment at:

toolstorepy_workspace/.venv

It installs the MCP runtime there so the generated server runs in isolation from your system Python environment.

Sanity check​

toolstorepy --help

Use it as a library​

from toolstorepy.orchestrator import ToolStorePy

toolstore = ToolStorePy(
    workspace="toolstorepy_workspace",
    install_requirements=True,
    host="0.0.0.0",
    port=9090,
    llm_scan=True,
    llm_model="claude-sonnet-4-6",
    verbose=True,
)

output_path = toolstore.build(
    queries="queries.json",
    index="core-tools",
)

print(output_path)

About index authoring tools​

The packaged CLI consumes indexes through --index or --index-url. If you are authoring your own index, prefer a JSON manifest as the source format. The current reference repository still includes legacy script names such as:

python vector_db_creation/toon_file_creation.py
python vector_db_creation/embed_toon.py

That workflow is documented in Creating Custom Tool Indexes.