> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mycelium-ai.co/llms.txt
> Use this file to discover all available pages before exploring further.

# Python

> Python access to the substrate via in-repo skills

The Python surface ships in the open-source [ai-brain-starter](https://github.com/mycelium-hq/ai-brain-starter) repo. The skills under `skills/` are runnable Python: each one reads the vault, calls deterministic services, and writes typed entries.

## Install

```bash theme={null}
git clone https://github.com/mycelium-hq/ai-brain-starter ~/Documents/MyVault
cd ~/Documents/MyVault
./bootstrap.sh
```

After install, the skills register automatically with Claude Code. Run them via slash commands (`/journal`, `/graphify`, `/diagnose`) from any session.

## Direct import

For embedding the substrate in another Python project, the typed-memory schemas, the resolver, and the graph utilities are importable from the repo's `lib/` and `schemas/` directories. A standalone Python SDK package (PyPI) is on the roadmap; until then, vendor the modules you need.

## Standalone SDK status

A standalone PyPI package is planned. Goal: install via `pip install mycelium-sdk` and call the runtime API from any Python codebase without cloning the substrate.

Until that lands, the recommended path is:

* For LLM-driven work inside Claude Code, use the slash-command skills.
* For programmatic access from outside Claude Code, call the runtime's HTTP API directly (see [/api-reference/runtime](/api-reference/runtime)).

## Examples

Skill examples and reference implementations live under `skills/<skill-name>/SKILL.md` in the substrate repo. The `synth-pr-to-sop` skill is a good reference for the typed-memory write path.