@spiderbrain/create

@spiderbrain/create

Give a repo its committed SpiderBrain understanding layer. Fetches your scored brain from the SpiderBrain API and writes a source-free .spiderbrain/ folder plus an AGENTS.md block, so any AI coding agent that later touches the repo reads it with zero setup.

This is the produce side. To read a committed folder, see @spiderbrain/read. Most people use the spiderbrain umbrella command instead of either directly.

Use

export SPIDERBRAIN_API_KEY=sb_live_...     # get one at https://spiderbrain.ai/dashboard?tab=keys
npx @spiderbrain/create --brain my-project

Writes .spiderbrain/ and merges a SpiderBrain block into AGENTS.md. Commit both.

spiderbrain-create [--brain <name|id>] [--root .] [--private] [--agents AGENTS.md]

• --brain — which cloud brain to use (default: the only one on your account).
• --root — the repo directory to write into (default: current directory).
• --private — include the weighted scores. For a PRIVATE in-repo brain only; a public brain must never carry them.
• --agents — the agent context file to merge into (default AGENTS.md).

What it writes

A source-free understanding set, documented in the format spec: manifest.json, structure.ndjson, and an AGENTS.block.md merged into your AGENTS.md. The public variant contains file paths, structure, and an edge-derivable blast radius only, never your source and never the scoring weights.

License

MIT.