/ Directory / Playground / JDocMunch MCP
← All servers ↗ GitHub
● Community jgravelle ⚡ Instant

JDocMunch MCP

by jgravelle · jgravelle/jdocmunch-mcp

Documentation retrieval that doesn't drown your context — structured section index lets Claude pull just the section it needs.

JDocMunch is a documentation MCP optimized for token efficiency. It pre-indexes docs into a navigable section tree (TOC + anchors), so Claude can pull a single section, follow cross-refs, and never dump 50KB of unrelated reference material. Designed for the 'I'm reading the AWS SDK docs and Claude keeps fetching the entire page' problem.

▶ Watch demo 📦 Install 💡 Use Cases

Why use it

Key features

Live Demo

What it looks like in practice

jdocmunch-mcp.replay ▶ ready
0/0

Install

Pick your client

~/Library/Application Support/Claude/claude_desktop_config.json  · Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "jdocmunch-mcp": {
      "command": "uvx",
      "args": [
        "jdocmunch-mcp"
      ]
    }
  }
}

Open Claude Desktop → Settings → Developer → Edit Config. Restart after saving.

~/.cursor/mcp.json · .cursor/mcp.json
{
  "mcpServers": {
    "jdocmunch-mcp": {
      "command": "uvx",
      "args": [
        "jdocmunch-mcp"
      ]
    }
  }
}

Cursor uses the same mcpServers schema as Claude Desktop. Project config wins over global.

VS Code → Cline → MCP Servers → Edit
{
  "mcpServers": {
    "jdocmunch-mcp": {
      "command": "uvx",
      "args": [
        "jdocmunch-mcp"
      ]
    }
  }
}

Click the MCP Servers icon in the Cline sidebar, then "Edit Configuration".

~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "jdocmunch-mcp": {
      "command": "uvx",
      "args": [
        "jdocmunch-mcp"
      ]
    }
  }
}

Same shape as Claude Desktop. Restart Windsurf to pick up changes.

~/.continue/config.json
{
  "mcpServers": [
    {
      "name": "jdocmunch-mcp",
      "command": "uvx",
      "args": [
        "jdocmunch-mcp"
      ]
    }
  ]
}

Continue uses an array of server objects rather than a map.

~/.config/zed/settings.json
{
  "context_servers": {
    "jdocmunch-mcp": {
      "command": {
        "path": "uvx",
        "args": [
          "jdocmunch-mcp"
        ]
      }
    }
  }
}

Add to context_servers. Zed hot-reloads on save.

claude mcp add jdocmunch-mcp -- uvx jdocmunch-mcp

One-liner. Verify with claude mcp list. Remove with claude mcp remove.

Use Cases

Real-world ways to use JDocMunch MCP

Pull just the API reference section you need from a giant doc site

👤 Devs working with cloud SDKs ⏱ ~10 min beginner

When to use: AWS SDK docs are 200 pages and Claude wants to fetch them all.

Prerequisites
  • Doc URL or local doc tree — First call: index the doc; subsequent calls hit the cache
Flow
  1. Index
    Use jdocmunch. Index the AWS S3 SDK docs at the URL. Show me top-level sections.✓ Copied
    → TOC tree returned, no full content yet
  2. Drill
    Show only the section on multipart upload — code samples included, intro skipped.✓ Copied
    → Targeted section content
  3. Follow
    Now follow the link to 'Aborting incomplete uploads' from that section.✓ Copied
    → Linked section pulled

Outcome: Answer in 2k tokens instead of 80k.

Pitfalls
  • Doc structure has weird anchors (PDF-only) — Pre-convert to Markdown first; jdocmunch parses Markdown best
Combine with: filesystem

Compare API doc sections across two SDK versions

👤 Devs upgrading dependencies ⏱ ~20 min intermediate

When to use: You're moving from v3 to v4; you want a precise diff of what changed.

Flow
  1. Index both
    Index doc-v3 and doc-v4. List sections that exist only in v4 or have changed.✓ Copied
    → Diff list of sections
  2. Drill into changed
    Show changes in the 'Authentication' section between v3 and v4.✓ Copied
    → Per-section diff

Outcome: Migration guide built from the docs themselves.

Pitfalls
  • Section IDs change between versions — Match by heading text instead of anchor

Combinations

Pair with other MCPs for X10 leverage

jdocmunch-mcp + filesystem

Save retrieved sections to project notes

Pull the multipart-upload section and save it to /docs/notes/s3-multipart.md.✓ Copied
jdocmunch-mcp + git-mcp-idosal

Pair docs with source-of-truth source code

Read the doc section, then GitMCP the actual SDK source for that method.✓ Copied

Tools

What this MCP exposes

ToolInputsWhen to callCost
index url_or_path First time on a doc set 1 fetch + parse
list_sections doc_id, depth? TOC navigation free (cached)
get_section doc_id, section_path Pull one section free (cached)
follow_link doc_id, from_section, link_target Chase cross-references free (cached)

Cost & Limits

What this costs to run

API quota
Per host's robots policy
Tokens per call
100–800 with structured retrieval
Monetary
Free
Tip
Index once, query many — cache makes follow-ups cheap

Security

Permissions, secrets, blast radius

Minimum scopes: Outbound HTTP for indexing
Credential storage: None for public docs
Data egress: Whatever doc host you index
Never grant: Indexing of private internal docs without auth control

Troubleshooting

Common errors and fixes

Index parse fails on JS-heavy doc site

Some sites need a headless browser; check if a static export exists

Sections missing after re-index

Doc structure changed; bust the cache and re-index

uvx can't find binary

PATH issue under MCP client; use explicit uvx jdocmunch-mcp

Alternatives

JDocMunch MCP vs others

AlternativeWhen to use it insteadTradeoff
Context7Curated library docs surface — known catalogLimited to indexed libraries
GitMCPRepo-as-docs (README + /docs)GitMCP doesn't structure-index large doc sites

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills