jMunch LLC
jMunch LLC develops local-first MCP retrieval tools that let AI coding agents retrieve precise code symbols, documentation sections, and data/schema information instead of reading whole files. Its stated goal is to reduce context-window use, latency, and AI token costs while preserving precise access to source material.
At a Glance
- Solo developers and freelancers
- Small software-engineering teams
- Enterprise engineering and platform teams
- Organizations using Claude Code, Cursor, Gemini, Windsurf, Codex, Antigravity, or other MCP clients
- +2 more
AI Tools by jMunch LLC
(1)
jCodeMunch
MCP Server for Symbol Retrieval
Discussions
No discussions yet
Be the first to start a discussion about jMunch LLC
Latest News
jCodeMunch v1.108.318 released with workflow safeguards, competitive benchmark infrastructure, watcher/index reconciliation, and tree-sitter installation diagnostics.
jCodeMunch v1.108.317 released with CI harness and dispatched publishing-workflow updates.
jCodeMunch v1.108.316 released with measurement/reporting changes.
PyPI package documentation reported reproducible 96.5% average token reduction against a grep-and-read agent across 15 task-runs on three repositories.
Products & Services
An MCP server for source-code exploration. It parses code with tree-sitter, builds a persistent symbol index, and retrieves exact functions, classes, methods, constants, outlines, and structural relationships by symbol rather than loading entire files.
A companion MCP server for indexing documentation and retrieving exact sections, headings, tables of contents, OpenAPI material, and related documentation content without loading complete documents.
A data-retrieval MCP server for indexing spreadsheets, databases, and data files and answering questions about schemas, columns, matching rows, joins, aggregates, correlations, and data-quality hotspots.
The open Munch Retrieval Interface specification used by jCodeMunch, jDocMunch, and jDataMunch to define deterministic payload shapes, token budgets, and stable tool contracts.
Market Position
jCodeMunch positions itself as a symbol-accurate retrieval scalpel rather than a file-oriented repository map. Relative to Aider RepoMap and standalone RepoMapper, it emphasizes exact symbol retrieval and MCP portability; relative to semantic code-search products such as Greptile and GrepAI, it emphasizes deterministic AST parsing, byte-offset precision, local-first operation, and low retrieval cost. Its key differentiator is surgical retrieval for agents that already know or can discover the relevant symbol, while graph- and embedding-based competitors can be stronger for broad intent discovery or cross-file context.
Leadership
Founders
J. Gravelle
Independent AI developer and the maintainer of jCodeMunch. Gravelle previously created AutoGroq, PocketGroq, Groqqle, and other tools in the Groq/AI-agent ecosystem; GitHub identifies Gravelle as the owner and primary maintainer of the jCodeMunch repository.
Executive Team
J. Gravelle
Founder, developer, and maintainer
Independent AI developer; GitHub owner and primary maintainer of jCodeMunch. Previous projects include AutoGroq, PocketGroq, and Groqqle.
Founding Story
The project began as a solo effort to address the waste created when AI coding agents repeatedly open and read entire source files. The initial vision was to index a repository once, parse its symbols, and let MCP-compatible agents retrieve only the exact functions, classes, methods, or other structures needed for a task.
Business Model
Revenue Model
One-time commercial software-license purchases; personal/non-commercial use is free under the repository license. Commercial licensing covers individual developers, small teams, and organization-wide deployments, with optional trio bundles for jCodeMunch, jDocMunch, and jDataMunch.
Pricing Tiers
Commercial use for one developer.
Commercial use for up to five developers.
Organization-wide internal deployment.
Bundle of jCodeMunch, jDocMunch, and jDataMunch for one developer.
Bundle for up to five developers.
Organization-wide bundle of all three retrieval tools.
Target Markets
- Solo developers and freelancers
- Small software-engineering teams
- Enterprise engineering and platform teams
- Organizations using Claude Code, Cursor, Gemini, Windsurf, Codex, Antigravity, or other MCP clients
- Open-source maintainers
- Teams managing AI-token budgets and large codebases
- AI-assisted development in large or unfamiliar repositories
- Repository onboarding and architecture exploration
- Finding and retrieving one exact function, class, or method for an agent task
- Authentication-flow tracing and dependency-injection analysis
- Refactoring impact analysis and blast-radius checking
- Cross-file importer/reference and dead-code analysis
- The official site cites community use and GitHub stars from engineers associated with Microsoft, Anthropic, OpenAI, NASA, IBM, Intel, Red Hat, Shopify, Siemens, Epic Games, Alibaba, ByteDance, and Stanford; these are presented as GitHub-star/community signals rather than named paying customers.