
Tokennuke
This is a code indexing server that gives Claude surgical access to large codebases without burning context tokens. It uses tree-sitter to parse 10 languages into ASTs, then stores symbols in SQLite with FTS5 and vector embeddings for hybrid search. You get 15 tools including get_symbol for O(1) byte-offset retrieval (grab a 200-byte function from a 40KB file), search_symbols with semantic and keyword matching, call graph traversal with get_callers and get_callees, and index_repo to pull down any GitHub or GitLab repository without needing git installed. Incremental indexing via SHA-256 hashing means re-indexing after one file change takes milliseconds. Reach for this when you're building agents that need to navigate unfamiliar codebases or doing refactoring work where you need to trace dependencies before moving code around.
- tools
- 15
- transport
- STDIO
- last updated
- Mar 2, 2026
Tools
15 toolsPublic tool metadata for what this MCP can expose to an agent.
index_folderIndex a local directory (incremental, SHA-256 based)
Index a local directory (incremental, SHA-256 based)
No parameter schema in public metadata yet.
index_repoIndex a GitHub/GitLab repo (tarball download, no git needed)
Index a GitHub/GitLab repo (tarball download, no git needed)
No parameter schema in public metadata yet.
list_reposList all indexed repositories with stats
List all indexed repositories with stats
No parameter schema in public metadata yet.
invalidate_cacheForce re-index a repository
Force re-index a repository
No parameter schema in public metadata yet.
file_treeGet directory tree with file counts
Get directory tree with file counts
No parameter schema in public metadata yet.
file_outlineList symbols in a single file
List symbols in a single file
No parameter schema in public metadata yet.
repo_outlineList all symbols in repo (summary)
List all symbols in repo (summary)
No parameter schema in public metadata yet.
get_symbolGet full source of one symbol (O(1) byte seek)
Get full source of one symbol (O(1) byte seek)
No parameter schema in public metadata yet.
get_symbolsBatch get multiple symbols
Batch get multiple symbols
No parameter schema in public metadata yet.
search_symbolsHybrid search (FTS5 + vector RRF)
Hybrid search (FTS5 + vector RRF)
No parameter schema in public metadata yet.
search_textFull-text search in file contents
Full-text search in file contents
No parameter schema in public metadata yet.
get_calleesWhat does this function call?
What does this function call?
No parameter schema in public metadata yet.
get_callersWho calls this function?
Who calls this function?
No parameter schema in public metadata yet.
diff_symbolsWhat changed since last index? (PR review)
What changed since last index? (PR review)
No parameter schema in public metadata yet.
dependency_mapWhat does this file depend on? What depends on it?
What does this file depend on? What depends on it?
No parameter schema in public metadata yet.
Sponsors
Readme
TokenNuke
Intelligent code indexing MCP server. 15 tools, 10 languages, tree-sitter AST extraction, hybrid search (FTS5 + vector), call graphs, remote repo indexing, incremental indexing.
Save 99% of tokens — get exact function source via byte-offset seek instead of reading entire files.
Formerly
codemunch-pro. Same code, better name.
Install
pip install tokennuke
Quick Start
Claude Desktop / Cline
Add to your MCP client config:
{
"mcpServers": {
"tokennuke": {
"command": "tokennuke"
}
}
}
HTTP Server
tokennuke --transport streamable-http --port 5002
15 MCP Tools
Details
- Repository
- github.com/bigjai/tokennuke
- Category
- Search & Web Crawling
- Package
tokennuke- Transport
- STDIO
- Registry
- MCP registry, active