jCodeMunch indexes your codebases and jDocMunch indexes your documentation — two MCP servers built for surgical AI retrieval. No more flooding context windows with thousands of lines of unrelated files or docs.
A production .NET 10 API suite with multiple service endpoints, authorization middleware, health checks, FTP/email integrations, and data models — used verbatim for this benchmark.
Compare a standard agent's file-based exploration vs. jCodeMunch symbol retrieval for the query: "How are survey submissions handled?"
Short version: RepoMapper is a ranked repository “map” (great for orientation and “what matters?”), while jCodeMunch is symbol-accurate retrieval (great for “show me the exact code” with tiny token spend). They overlap, but they’re optimized for different jobs.
{file_path}::{qualified_name}#{kind} lets an agent “bookmark” code reliably across sessions.Antigravity uses a standard MCP config file — setup takes about a minute.
pip install git+https://github.com/jgravelle/jcodemunch-mcp.gitmcp_config.json{
"mcpServers": {
"jcodemunch": {
"command": "jcodemunch-mcp",
"env": {
"GITHUB_TOKEN": "ghp_...",
"ANTHROPIC_API_KEY": "sk-ant-..."
}
}
}
}
ANTHROPIC_API_KEY enables AI-generated symbol summaries; GITHUB_TOKEN raises GitHub API rate limits and unlocks private repos.jCodeMunch munches code. jDocMunch munches documentation — the same surgical retrieval approach, applied to Markdown, READMEs, specs, and any text-based docs in your repo.
search_sections(“auth flow”) — one call, the right section, nothing else.
pip install git+https://github.com/jgravelle/jdocmunch-mcp.git
—
github.com/jgravelle/jdocmunch-mcp ↗
These are verbatim results from the jCodeMunch MCP server querying the indexed Bakery Deli Survey codebase.
Each file read floods the context window. jCodeMunch retrieves only the symbol requested.
| File | Lines | Tokens (Traditional) | Tokens (jCodeMunch) | Savings |
|---|---|---|---|---|
| Program.cs | 1,031 | 8,972 | ~0 (not needed) | 100% |
| SurveyService.cs | 520 | 6,102 | ~340 (one method) | 94.4% |
| AuthorizationService.cs | 320 | 2,840 | ~110 (one helper) | 96.1% |
Calculating the actual dollar impact of context-window waste on a 141K token codebase.
Scale to multiple projects and more queries per day, the savings multiply accordingly.
A pre-built symbol index lets the MCP server answer code queries in milliseconds with surgical precision.
Run index_code_folder(path) to index code symbols, or index_doc_local(path) (jDocMunch) to index documentation sections. Both build a persistent local index — happens once per project.
Instead of reading files, the AI calls search_symbols(query) or get_symbol(id). The MCP server performs semantic + keyword search against the index in milliseconds.
Only the matching symbol's source code and metadata is returned — not the surrounding file, not unrelated classes. A 6,000-token file read is replaced by a 400-token symbol pull.
Choose a single-product license for code or docs, or get both in a Munch Duo bundle. All licenses are commercial-use licenses for the specified tier.