
Codeseeker
Index your codebase as a knowledge graph so agents semantically search symbols, flows, and relationships instead of blind grep.
Overview
CodeSeeker is a MCP server for the Build phase that provides graph-powered semantic code search and a knowledge graph for AI assistants.
What is this MCP server?
- Graph-powered code intelligence with semantic search over your project
- Knowledge graph representation for relationships between code entities
- npm package codeseeker v2.0.1 with stdio MCP transport
- Designed for AI assistants pairing with local repo checkout
- Published version 2.0.1 on npm registry identifier codeseeker
- Transport: stdio
- Repository: github.com/jghiringhelli/codeseeker
Community signal: 18 GitHub stars.
What problem does it solve?
Large codebases overwhelm agents with irrelevant files because keyword search misses conceptual links between components.
Who is it for?
Indie developers on non-trivial repos who want Cursor or Claude Code to navigate architecture and dependencies with graph-aware search.
Skip if: Tiny single-file prototypes or teams unwilling to run local indexing tooling beside the repo.
What do I get? / Deliverables
After installing the codeseeker npm MCP via stdio, assistants can query semantic code intelligence grounded in a project knowledge graph.
- stdio MCP CodeSeeker process wired to your agent
- Semantic search over project symbols and flows
- Knowledge-graph-backed answers about code relationships
Recommended MCP Servers
Journey fit
Code intelligence peaks in Build but accelerates Ship review and Operate debugging when the graph stays current. Agent-tooling captures MCP servers that augment how assistants navigate and reason about repository structure.
How it compares
Repo-local code graph MCP, not hosted cross-session memory like 0Latency.
Common Questions / FAQ
Who is CodeSeeker for?
Solo builders and small teams whose AI assistants need semantic code search and relationship-aware answers across a real codebase.
When should I use CodeSeeker?
Use it during Build feature work and when reviewing or debugging complex modules where graph context beats plain text search.
How do I add CodeSeeker to my agent?
Install the npm package codeseeker (v2.0.1), configure stdio MCP in your client, and point it at your indexed project per the GitHub README.