jgravelle

Jflowmunch — AI skill for Claude Code

AI community

Live flow diagram of what an AI coding agent intends to do, rendered from a watched JSON file.

How to install Jflowmunch

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

What Jflowmunch does

Live flow diagram of what an AI coding agent intends to do, rendered from a watched JSON file.

Alternatives in AI

Claude Flow — An enterprise-grade AI orchestration platform that revolutionizes how developers build with AI 22.2k ★
Architecture Diagram Generator — Generate beautiful dark-themed system architecture diagrams as standalone HTML/SVG files 7k ★
Flow Define — Multi-AI requirements scoping using Codex and Gemini CLIs (Double Diamond Define phase) 2.8k ★

README

jFlowMunch

A live flow diagram of what an AI coding agent intends to do, in a browser tab you keep open beside the conversation.

The agent describes a proposed change in prose. Alongside the prose it writes a small JSON flow document to disk. A local server watches that file and pushes a rendered flowchart to the tab. You glance at the diagram to confirm the agent understood the request before you approve it.

jFlowMunch is part of the jMunch suite (jCodeMunch, jDocMunch, jDataMunch). The other tools reduce the tokens an agent spends understanding a codebase. This one reduces the time a human spends understanding the agent.

Install

pip install jflowmunch

Standard library only; no runtime dependencies. Python 3.10 or newer.

Use

Start the server in the repo you are working in:

jflowmunch serve

It prints a URL (default `http://127.0.0.1:7431/`). Open it. The page says which file it is waiting for. That file lives under your home directory, never in the project:

~/.jFlowMunch/-<8 hex of its path>/current.json

`jflowmunch where` prints the path for the current directory. Every command derives it the same way from the directory it runs in, so the server, the hooks and the agent agree without configuration. Set `JFLOWMUNCH_HOME` to move the base directory; pass `--watch PATH` to override it for one command.

Nothing is written inside the repository. A project you point jFlowMunch at needs no ignore entry, no config file, and no line in its CLAUDE.md.

Tell your agent to write flow documents. Nothing in the target repo has to change for that. Add two hooks to your user-level Claude Code settings (`~/.claude/settings.json`):

{
  "permissions": {
    "allow": ["Edit(~/.jFlowMunch/**)"]
  },
  "hooks": {
    "UserPromptSubmit": [
      {
        "hooks": [{ "type": "command", "command": "jflowmunch prompt" }]
      }
    ],
    "PostToolUse": [
      {
        "matcher": "Edit|Write|MultiEdit",
        "hooks":