claude-tap
claude-tap is a local proxy and trace viewer that shows what AI coding agents send and receive, for developers debugging agent behavior.
What it is good for
- Seeing the exact prompts a coding agent sends to the model
- Debugging why a coding agent behaved a certain way
- Comparing how context changes between agent requests
- Checking token usage and tool calls in an agent session
At a glance
- Who it is for
- Setup difficulty
- Technical ยท Guided command-line setup
Level 4 of 5. You copy and paste commands into a terminal by following the instructions. Comfortable territory for vibe coders.
- Cost
- Free and open source
- Type
- Agent app
- Tasks
- Works with
- Claude Code, Codex CLI, Gemini CLI, Cursor CLI, OpenCode, Hermes Agent
How to get started with claude-tap
- Install Python 3.11 or newer, plus the coding agent you want to trace.
- Install the tool with
uv tool install claude-taporpip install claude-tap. - Run
claude-tapin place ofclaudeto launch Claude Code with the live browser viewer. - Pass client flags after
--, for exampleclaude-tap -- --model claude-opus-4-6. - Use
claude-tap --tap-client cursorto trace Cursor CLI.
From the project's own documentation. If a step does not match what you see, the official pages have the latest instructions.
Before you try it
- Requires Python 3.11+ and the client you want to trace.
- Common auth headers are replaced with *** in recordings, but request and response bodies and URLs are not redacted.
- Traces are stored locally and no hosted dashboard is required.
Momentum
- Rank this week
- Not ranked yet
- Best rank
- None yet
- GitHub stars
- 3,344
GitHub numbers as of Nov 16, 2026. Updated weekly.
Official links
Questions
- Is claude-tap free?
- Yes. It is free and open source (MIT license).
- Do I need to know how to code to use claude-tap?
- You do not need to write code, but setup means copying and pasting commands into a terminal by following its instructions.
- What does claude-tap work with?
- According to its own documentation, it works with Claude Code, Codex CLI, Gemini CLI, Cursor CLI, OpenCode, and Hermes Agent.
- How do I get started with claude-tap?
- 1. Install Python 3.11 or newer, plus the coding agent you want to trace. 2. Install the tool with uv tool install claude-tap or pip install claude-tap. 3. Run claude-tap in place of claude to launch Claude Code with the live browser viewer. 4. Pass client flags after --, for example claude-tap -- --model claude-opus-4-6. 5. Use claude-tap --tap-client cursor to trace Cursor CLI.
- What should I know before trying claude-tap?
- Requires Python 3.11+ and the client you want to trace. Common auth headers are replaced with *** in recordings, but request and response bodies and URLs are not redacted. Traces are stored locally and no hosted dashboard is required.
- What is claude-tap good for?
- Seeing the exact prompts a coding agent sends to the model, debugging why a coding agent behaved a certain way, comparing how context changes between agent requests, and checking token usage and tool calls in an agent session.
Similar agents to claude-tap
- agent-trace
Records and replays coding agent sessions to check what the agent did, rather than inspecting raw API traffic.
- Nasiko
Finds coding agents on your machine, tracks their costs, and routes them to models, rather than inspecting requests.
- OpenHands
Runs coding agents and automations from a control center instead of inspecting their traffic.
- agentmemory
Gives coding agents persistent memory across sessions instead of showing what they sent to the model.
- Pi
A desktop app to watch and coordinate Pi coding agents, not a proxy for several agent clients.
- opencode
opencode is an open source coding agent that writes and changes code in your project for developers.