Get started

Install BrainLayer

pip install brainlayer

*Requires Python 3.10+. Installs the CLI, MCP server, and all dependencies.

Index your conversations

Point BrainLayer at your Claude Code sessions directory. It will parse JSONL transcripts, chunk content with AST-aware splitting, and generate 1024-dim embeddings.

brainlayer index

*First run indexes all sessions. Subsequent runs only process new/changed files.

Configure MCP

Add BrainLayer to your Claude Code MCP configuration so it's available in every session.

// ~/.claude/settings.json
{
  "mcpServers": {
    "brainlayer": {
      "command": "brainlayer-mcp"
    }
  }
}

Run your first search

Search your development history using natural language. Results include content, enrichment metadata, and relevance scores.

brainlayer search "how did I implement auth"

*Example output: "Found 12 results across 3 projects. Top: JWT middleware in shared/auth.ts (score: 0.89, importance: 8, intent: implementing)"

Enrich your chunks (optional)

Run a local LLM to add 10-field metadata to every chunk: summary, tags, importance, intent, and more. Requires Ollama with GLM-4.7-Flash or an MLX-compatible model.

brainlayer enrich

*This is optional but dramatically improves search quality. Processing runs locally — no data leaves your machine.