Skip to content

OpenClaw

OpenClaw is a local-first AI assistant that connects to 22+ messaging platforms and routes conversations through configurable LLM backends. It supports MCP natively — adding NestWeaver gives every conversation structural code intelligence without reading raw files.

Terminal window
openclaw mcp add nestweaver \
--command nestweaver \
--arg mcp \
--arg --db \
--arg /path/to/your.lbug

OpenClaw probes the server on save to verify the connection. Use --no-probe if NestWeaver isn’t running yet.

Terminal window
openclaw mcp set nestweaver '{"command":"nestweaver","args":["mcp","--db","/path/to/your.lbug"]}'

Add to ~/.openclaw/openclaw.json under mcp.servers:

{
"mcp": {
"servers": {
"nestweaver": {
"command": "nestweaver",
"args": ["mcp", "--db", "/path/to/your.lbug"]
}
}
}
}

If the OpenClaw gateway is running, open http://localhost:18789/mcp in your browser and add the server through the UI.

Terminal window
openclaw mcp doctor nestweaver --probe

This runs a static config check and a live connection test. To list all discovered tools:

Terminal window
openclaw mcp probe nestweaver --json

NestWeaver exposes 40 tools by default. To limit which tools OpenClaw sees:

Terminal window
openclaw mcp tools nestweaver --include 'brain_context,brain_search,read_symbols,blast_radius,brain_impact'

Or exclude specific tools:

Terminal window
openclaw mcp tools nestweaver --exclude 'prune_stale,set_extension'

Clear filters to restore all tools:

Terminal window
openclaw mcp tools nestweaver --clear

You can also set filters in the config JSON:

{
"nestweaver": {
"command": "nestweaver",
"args": ["mcp", "--db", "/path/to/your.lbug"],
"toolFilter": {
"include": ["brain_context", "brain_search", "read_symbols", "blast_radius"],
"exclude": []
}
}
}

Alternatively, use NestWeaver’s built-in lite mode for a minimal 6-tool set:

{
"nestweaver": {
"command": "nestweaver",
"args": ["mcp", "--lite", "--db", "/path/to/your.lbug"]
}
}

Without NestWeaver, OpenClaw navigates code using grep and file reads — multiple round-trips to find relevant symbols, each consuming tokens and context window space. With NestWeaver:

  • Single-call context retrievalbrain_context returns ranked, token-budget-aware results instead of grep-then-read chains
  • Structural queries — “What calls this function?” and “What’s the blast radius?” answered from the graph, not by reading 10+ files
  • Persistent codebase memory — The graph persists across sessions. No re-discovering project structure every conversation.
  • Reduced compaction — Less raw code loaded into context means OpenClaw’s context compaction triggers less often and loses less information
Terminal window
openclaw mcp list # List all configured servers
openclaw mcp show nestweaver --json # Show server config
openclaw mcp status --verbose # All servers with transport details
openclaw mcp reload # Dispose cached MCP runtimes

To keep the config but skip NestWeaver at startup:

{
"nestweaver": {
"command": "nestweaver",
"args": ["mcp", "--db", "/path/to/your.lbug"],
"enabled": false
}
}