Spiderbrain MCP server
{
"mcpServers": {
"spiderbrain": {
"command": "npx",
"args": [
"-y",
"spiderbrain@latest"
],
"env": {
"SPIDERBRAIN_API_KEY": "<spiderbrain_api_key>"
}
}
}
}{
"mcpServers": {
"spiderbrain": {
"command": "npx",
"args": [
"-y",
"spiderbrain@latest"
],
"env": {
"SPIDERBRAIN_API_KEY": "<spiderbrain_api_key>"
}
}
}
}{
"mcpServers": {
"spiderbrain": {
"command": "npx",
"args": [
"-y",
"spiderbrain@latest"
],
"env": {
"SPIDERBRAIN_API_KEY": "<spiderbrain_api_key>"
}
}
}
}{
"servers": {
"spiderbrain": {
"command": "npx",
"args": [
"-y",
"spiderbrain@latest"
],
"env": {
"SPIDERBRAIN_API_KEY": "${input:spiderbrain_api_key}"
}
}
},
"inputs": [
{
"id": "spiderbrain_api_key",
"type": "promptString",
"description": "Spiderbrain api key",
"password": true
}
]
}{
"mcpServers": {
"spiderbrain": {
"command": "npx",
"args": [
"-y",
"spiderbrain@latest"
],
"env": {
"SPIDERBRAIN_API_KEY": "<spiderbrain_api_key>"
}
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsoninstall npx -y spiderbrainversion 0.2.3
What it does
SpiderBrain is an MCP server that reads a pre-computed `.spiderbrain/` understanding layer committed alongside a repository, exposing dependency structure, blast radius, and keystone files to AI coding agents. It is aimed at developers using agents like Claude Code or Cursor who want deterministic, session-stable answers about change impact without re-deriving structure each time.
- Query which files are affected by a change to a specific source file before committing it.
- Gate pull requests in CI by failing builds whose blast radius exceeds a configured file count.
- Identify load-bearing files in a codebase to understand architectural risk before refactoring.
Tools this server exposes
| Tool | Does |
|---|---|
| sb_blast | Returns the set of files reachable from a given source file when it changes. |
| sb_impact | Returns the blast radius of the current working diff. |
| sb_path | Shows how one file reaches another through the dependency graph. |
| sb_keystones | Lists the load-bearing files in the repository based on the understanding layer. |
| sb_map | Returns structural information about a file and what it touches. |
| sb_ask | Answers questions about the repository using the understanding layer, with cloud fallback for semantic queries when an API key is set. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.