Back to Apps

jCodeMunch
by jgravelle
Cut AI token costs by 95%+ with precise, symbol-level GitHub code retrieval via tree-sitter AST.
0 stars
Works in:CursorCodexWindsurf
Exposes:ToolsResources
What it does
jCodeMunch is a high-precision MCP server designed to stop AI agents from wasting tokens by reading entire files. It indexes codebases using tree-sitter AST parsing, allowing agents to retrieve only the specific functions, classes, or symbols they need with byte-level precision.
Tools
search_symbols: Find functions, classes, or constants using fuzzy matching.get_symbol_source: Retrieve the exact implementation of a specific symbol.get_file_outline: Get a structural map of a file before reading it.assemble_task_context: Orchestrate a multi-tool sequence to gather all context for a task in one call.plan_turn: Analyze a query to determine the most efficient retrieval route.get_blast_radius: Calculate the impact of changing a symbol across the codebase.find_dead_code: Identify unreachable symbols and files.
Installation
{
"mcpServers": {
"jcodemunch": {
"command": "uvx",
"args": ["jcodemunch-mcp"]
}
}
}
Supported hosts
Claude Desktop, Claude Code, Cursor, VS Code, Codex CLI, Continue, Windsurf
Quick install
uvx jcodemunch-mcpInformation
- Pricing
- freemium
- Published
- 7/1/2026
- stars
- 0
Categories
Choose your AI client and follow the steps below.
Codex
Add [mcp_servers.jcodemunch] block to ~/.codex/config.tomlCursor
mcp.json entry: { "name": "jcodemunch", "command": "uvx", "args": ["jcodemunch-mcp"] }Windsurf
mcp_config.json entryClaude Desktop
{
"mcpServers": {
"jcodemunch": {
"command": "uvx",
"args": ["jcodemunch-mcp"]
}
}
}





