codeseeker — Claude Code MCP server: what it does, how to install
Combines lexical and vector retrieval with an AST knowledge graph to trace dependencies and logic across codebases.
claude mcp add codeseeker -- npx -y codeseekerWhat codeseeker does
CodeSeeker merges BM25 lexical matching with 384-dimensional Xenova embeddings and RAPTOR directory summaries. During indexing, it parses AST imports to construct a dependency graph of calls and inheritance across files.
Incoming queries run through Reciprocal Rank Fusion, downranking test files and traversing structural graph edges. This surfaces indirect dependencies that basic text search skips.
Building the repository index requires between 30 seconds and 5 minutes upfront. Graph retrieval cannot function without completing this initial indexing step.
What it helps with
- locate token refresh mechanisms across distributed files
- trace import chains and class inheritance hierarchies
- suppress test file noise in architectural queries
- detect dead code by traversing AST relationship graphs
Example prompts for the agent
- Locate the authentication handler and list all modules that import it
- Find token verification logic and inspect its direct callers
How to install codeseeker
Platform CLI
claude mcp add codeseeker -- npx -y codeseekerPlatform CLI
codex mcp add codeseeker -- npx -y codeseekerConfig file
{
"mcpServers": {
"codeseeker": {
"command": "npx",
"args": [
"-y",
"codeseeker"
]
}
}
}Config file
{
"servers": {
"codeseeker": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"codeseeker"
]
}
}
}Platform CLI
gemini mcp add codeseeker npx -y codeseekerStep by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about codeseeker
How to install codeseeker in Claude Code?
Run claude mcp add codeseeker -- npx -y codeseeker in your terminal.
What does codeseeker do?
It provides code navigation for AI assistants by fusing BM25 keyword search, vector embeddings, and an AST dependency graph.
How long does initial indexing take?
Building the project index takes from 30 seconds to 5 minutes depending on codebase size.
