Agent Flow Visualization logo

Agent Flow Visualization

Organization
reason-machines
agent-flow-visualization

VS Code extension for real-time visualization of Claude Code agent orchestration as interactive node graphs

Overview

Publisherreason-machines
Repositorytrending-skills
Skill nameagent-flow-visualization
Stars
80
Forks
15
Bundled files
Instructions only
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by reason-machines on GitHub. Read the source before you install it.

Installation

Install the Agent Flow Visualization AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/reason-machines/trending-skills.git /tmp/trending-skills
mkdir -p .claude/skills
cp -r /tmp/trending-skills/skills/agent-flow-visualization .claude/skills/agent-flow-visualization
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Agent Flow Visualization in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Agent Flow Visualization on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Agent Flow Visualization is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Agent Flow

Skill by ara.so — Daily 2026 Skills collection.

Agent Flow is a VS Code extension that provides real-time visualization of Claude Code agent orchestration. It renders agent execution as an interactive node graph, showing tool calls, branching, subagent coordination, and timing — turning Claude Code's black-box execution into a transparent, debuggable flow.

Installation

Via VS Code Marketplace

  1. Open VS Code Extensions (Cmd+Shift+X)
  2. Search for Agent Flow by simon-p
  3. Click Install

Or install directly: marketplace.visualstudio.com/items?itemName=simon-p.agent-flow

Requirements

  • VS Code 1.85 or later
  • Claude Code CLI installed and accessible
  • Node.js (for Claude Code)

Quick Start

# 1. Open Command Palette
Cmd+Shift+P (Mac) / Ctrl+Shift+P (Win/Linux)

# 2. Run the command
> Agent Flow: Open Agent Flow

# 3. Start a Claude Code session in your workspace
# Agent Flow auto-detects it and begins streaming

Or use the keyboard shortcut:

  • Mac: Cmd+Alt+A
  • Win/Linux: Ctrl+Alt+A

Commands

CommandDescription
Agent Flow: Open Agent FlowOpen visualizer in current editor column
Agent Flow: Open Agent Flow to SideOpen in a side editor column
Agent Flow: Connect to Running AgentManually connect to a specific agent session
Agent Flow: Configure Claude Code HooksSet up Claude Code hooks for live event streaming

Configuration

Settings available in VS Code settings (settings.json):

jsonc
{
  // Path to a JSONL event log file to watch/replay
  "agentVisualizer.eventLogPath": "/path/to/agent-events.jsonl",

  // Auto-open the visualizer when an agent session starts
  "agentVisualizer.autoOpen": true,

  // Development server port (0 = production mode, use built assets)
  "agentVisualizer.devServerPort": 0
}

Auto-Open on Agent Start

jsonc
// settings.json
{
  "agentVisualizer.autoOpen": true
}

Claude Code Hooks Setup

Agent Flow uses Claude Code's hook system for zero-latency event streaming. Hooks are configured automatically on first open, but you can reconfigure manually.

Automatic Configuration

Run from Command Palette:

> Agent Flow: Configure Claude Code Hooks

Manual Hook Configuration

If you need to configure hooks manually, Agent Flow starts a local HTTP server that receives events. The hooks forward Claude Code lifecycle events (tool calls, responses, session start/end) to the extension.

Claude Code hooks are configured in ~/.claude/settings.json or project-level .claude/settings.json:

jsonc
{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": ".*",
        "hooks": [
          {
            "type": "command",
            "command": "curl -s -X POST http://localhost:PORT/hook -H 'Content-Type: application/json' -d @-"
          }
        ]
      }
    ],
    "PostToolUse": [
      {
        "matcher": ".*",
        "hooks": [
          {
            "type": "command",
            "command": "curl -s -X POST http://localhost:PORT/hook -H 'Content-Type: application/json' -d @-"
          }
        ]
      }
    ]
  }
}

Agent Flow manages this configuration automatically — prefer using the command palette command.

JSONL Event Log Mode

For replaying past sessions or watching log files generated outside VS Code:

jsonc
// settings.json
{
  "agentVisualizer.eventLogPath": "${workspaceFolder}/logs/agent-session.jsonl"
}

Agent Flow tails the file and visualizes events as they arrive. Use this for:

  • Post-hoc debugging of agent runs
  • Sharing agent execution recordings with teammates
  • CI/CD pipeline agent monitoring

JSONL Event Format

Each line in the log file is a JSON event object:

jsonl
{"type":"session_start","sessionId":"abc123","timestamp":"2026-03-21T10:00:00Z","model":"claude-opus-4-5"}
{"type":"tool_use","sessionId":"abc123","toolName":"bash","input":{"command":"ls -la"},"timestamp":"2026-03-21T10:00:01Z"}
{"type":"tool_result","sessionId":"abc123","toolName":"bash","output":"total 48\n...","timestamp":"2026-03-21T10:00:02Z"}
{"type":"message","sessionId":"abc123","role":"assistant","content":"I can see the files...","timestamp":"2026-03-21T10:00:03Z"}

Multi-Session Support

Agent Flow tracks multiple concurrent Claude Code sessions with tabs. Each session gets its own visualization canvas.

# Start multiple Claude Code sessions in different terminals
# Each appears as a separate tab in Agent Flow
# Click tabs to switch between session graphs

Interactive Canvas Features

Navigation

  • Pan: Click and drag on empty canvas
  • Zoom: Scroll wheel / pinch gesture
  • Select node: Click any agent or tool call node to inspect details
  • Reset view: Double-click empty canvas

Node Types in the Graph

Node TypeVisualDescription
Agent/SubagentCircleClaude instance making decisions
Tool CallRectangleIndividual tool invocation (bash, read_file, etc.)
Tool ResultRectangle (dashed)Output returned from tool
BranchDiamondDecision point spawning subagents
ReturnArrowSubagent returning result to parent

Panels

  • Timeline: Chronological view of all events with durations
  • Transcript: Full message history between agent and tools
  • File Heatmap: Which files received the most attention

Contributing / Development Setup

bash
git clone https://github.com/patoles/agent-flow
cd agent-flow
npm install

# Build the extension
npm run build

# Watch mode for development
npm run watch

# Run with dev server (hot reload)
# Set in settings: "agentVisualizer.devServerPort": 3000
npm run dev

Extension Structure

agent-flow/
├── src/
│   ├── extension.ts          # VS Code extension entry point
│   ├── hookServer.ts         # HTTP server receiving Claude Code hook events
│   ├── sessionManager.ts     # Manages multiple agent sessions
│   ├── webviewProvider.ts    # Webview panel management
│   └── logWatcher.ts         # JSONL file tail watcher
├── webview/
│   ├── src/
│   │   ├── App.tsx           # Main React app
│   │   ├── Graph.tsx         # Node graph canvas (likely D3 or React Flow)
│   │   ├── Timeline.tsx      # Timeline panel
│   │   └── Transcript.tsx    # Message transcript panel
│   └── package.json
└── package.json              # Extension manifest

Common Patterns

Pattern: Debug a Failing Agent Run

  1. Open Agent Flow (Cmd+Alt+A)
  2. Run your Claude Code command that's failing
  3. Watch the graph — look for:
    • Red/error nodes indicating failed tool calls
    • Unexpected branching
    • Loops (agent retrying the same tool)
  4. Click the failed tool node to inspect input/output
  5. Check the Timeline panel for which tool call took unexpectedly long

Pattern: Replay a Saved Session

bash
# Save Claude Code output to JSONL during a run
# (depends on your Claude Code version/config)
claude --output-format jsonl > session-$(date +%s).jsonl

# Point Agent Flow at it
jsonc
// .vscode/settings.json (project-level)
{
  "agentVisualizer.eventLogPath": "./logs/session-1234567890.jsonl"
}

Pattern: Auto-Visualize All Agent Sessions

jsonc
// settings.json
{
  "agentVisualizer.autoOpen": true
}

Now every time a Claude Code session starts in your workspace, Agent Flow opens automatically.

Pattern: Side-by-Side Coding and Visualization

# Open Agent Flow to side so you can code and watch simultaneously
Cmd+Shift+P > Agent Flow: Open Agent Flow to Side

This opens the visualizer in a split editor, keeping your code files accessible in the main editor group.

Troubleshooting

Agent Flow doesn't detect my Claude Code session

  1. Verify Claude Code is running in the same workspace folder
  2. Check that hooks are configured: run Agent Flow: Configure Claude Code Hooks
  3. Verify the hook server is running — look for "Agent Flow hook server listening" in the Output panel (View > Output > Agent Flow)
  4. Try Agent Flow: Connect to Running Agent to manually specify the session

Hooks aren't forwarding events

bash
# Test the hook server manually
curl -X POST http://localhost:PORT/hook \
  -H 'Content-Type: application/json' \
  -d '{"type":"test","sessionId":"test123"}'

Check VS Code Output panel for the correct PORT value.

Graph is empty / not updating

  1. Check agentVisualizer.eventLogPath — if set, Agent Flow reads from file instead of hooks
  2. Clear the setting if you want live hook-based streaming:
    jsonc
    { "agentVisualizer.eventLogPath": "" }
  3. Reload VS Code window (Cmd+Shift+P > Developer: Reload Window)

Extension not loading

bash
# Check VS Code version meets requirement
code --version
# Must be 1.85 or later

# Check extension is enabled
# Extensions panel > search "Agent Flow" > verify enabled

Multiple sessions showing in wrong tabs

Each session is identified by a unique session ID from Claude Code. If sessions are merging incorrectly, check that each claude process is started fresh (not reusing an existing session ID).

Resources

Frequently asked questions

What does the Agent Flow Visualization AI skill do?

VS Code extension for real-time visualization of Claude Code agent orchestration as interactive node graphs

Why use Agent Flow Visualization on TypingMind?

Because you install it once and use it with any model. Agent Flow Visualization is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Agent Flow Visualization in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/reason-machines/trending-skills/tree/main/skills/agent-flow-visualization. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Agent Flow Visualization?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Agent Flow Visualization?

As many as you like. As long as a model supports skills, you can use Agent Flow Visualization with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Agent Flow Visualization AI skill free?

It is published on GitHub by reason-machines. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇