jgravelle

Mcp Retrieval Spec

Name: Mcp Retrieval Spec
Author: jgravelle

Testing community

SOURCE https://github.com/jgravelle/mcp-retrieval-spec

Description

jMRI v1.0 — open specification for token-efficient context retrieval in MCP servers. SDKs, reference server, benchmark.

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

README

mcp-retrieval-spec

The jMRI (jMunch Retrieval Interface) specification — an open interface standard for token-efficient context retrieval in MCP servers.

What Is jMRI?

Agents that read whole files to answer specific questions waste 99% of their token budget. jMRI is a minimal interface for MCP servers that do retrieval right: index once, search by intent, retrieve exactly what you need.

Four operations. One response envelope. Two compliance levels.

**The problem in numbers:** A typical FastAPI codebase costs ~42,000 tokens to read naively. jMRI retrieval of the same answer costs ~480 tokens. At $3/1M tokens, that's $0.126 vs. $0.0014 per query. Across millions of queries, the savings are material.

The jMunch tools have saved billions of tokens across user sessions. This spec is the formal definition of what they do.

How It Works

Agent
  │
  ├─ discover()    → What knowledge sources are available?
  ├─ search(query) → Which symbols/sections are relevant? (IDs + summaries only)
  ├─ retrieve(id)  → Give me the exact source for this ID.
  └─ metadata(id?) → What would naive reading have cost?

Every response includes a `_meta` block with `tokens_saved` and `total_tokens_saved`. Agents can see exactly what they're saving on every call.

Spec

→ [SPEC.md](./SPEC.md)

The full jMRI v1.0 specification. Apache 2.0. Implement it however you want.

Reference Implementations

The spec is open. The best implementations are commercial.

Implementation	Domain	Stars	Install
jCodeMunch	Code (70+ languages)	1,500+	`uvx jcodemunch-mcp`
jDocMunch	Docs (MD, RST, HTML, notebooks)	135+	`uvx jdocmunch-mcp`
jDataMunch	Tabular data (CSV, Excel, Parquet, JSONL)	new	`uvx jdatamunch-mcp`