Skip to content

Latest commit

 

History

History
176 lines (117 loc) · 6.4 KB

README.md

File metadata and controls

176 lines (117 loc) · 6.4 KB

AgentClick

AgentClick icon

🧠 Human-in-the-loop review UI for autonomous AI agents

AgentClick is a skill-based plugin that adds a browser review layer to terminal agents.
Agents propose actions, users inspect and edit in a browser UI, then the agent continues.

npm version license npm downloads GitHub stars

AgentClick overview

⚙️ How It Works

Most agents still interact like this:

User → Terminal → Agent → Action

AgentClick adds a review step:

Agent proposes → Browser UI opens → User inspects and edits → Agent executes

This keeps the speed of autonomous agents while adding a human-in-the-loop safeguard before irreversible work.

🤖 Supported Agents

AgentClick works as a skill / plugin for modern AI agents. Any agent that can run local tools, send HTTP requests, and follow skill instructions can integrate with it.

  • Claude Code
  • Codex
  • OpenClaw
  • Custom tool-calling agents

🧩 What It's For

AgentClick extends the agent interaction into a browser UI for tasks like:

  • 📧 Email Review — browse inbox, edit drafts paragraph-by-paragraph, preference learning
  • 📋 Plan Review — inspect multi-step plans, inject constraints before execution
  • 🖥️ Code Review — GitHub-style diff viewer before any script runs
  • ✅ Action Approval — approve or reject external actions with reasons
  • 🔍 Trajectory Inspection — replay every decision, tool call, and error in a DAG
  • 🗳️ Selection Review — agent presents options, you choose
  • 🧠 Memory Management — control what the agent remembers across sessions

The goal: keep the speed of terminal agents, but add a real review layer before the agent commits to irreversible work.

✨ Why It Helps

  • Edit before execution — change the draft, command, or payload instead of only approve/reject.
  • Shared visual context — move from raw terminal text to a purpose-built UI.
  • Preference learning — feedback from review is persisted so the agent improves over time.
  • Framework-agnostic — anything that can POST JSON and poll an HTTP endpoint can use it.

🚀 Quick Start

npm install -g @harvenstar/agentclick
agentclick

Then open http://localhost:38173 in your browser.

For remote access from another device:

agentclick --remote

--remote automatically downloads and starts a Cloudflare Quick Tunnel. It prints a public HTTPS URL you can open on your phone or another machine.

🧠 Use It With An Agent

Option A — npm global install (AgentClick already running)

Start AgentClick first (agentclick), then tell your agent:

AgentClick is running at http://localhost:38173. Use it whenever you need a browser review UI — for email triage, risky commands, plans, or memory review.

Option B — working inside the AgentClick repo

If your agent can read local files, point it at the skill router:

Load SKILL.md from this repo, start AgentClick locally, and use it whenever you need a browser review UI instead of only terminal output.

The root SKILL.md routes to the right sub-skill automatically.

For OpenClaw in particular, use a stronger model with solid instruction-following, since the workflow is skill-based and depends on the agent following routing instructions reliably.

🧱 Skill Layout

AgentClick is built around a skill-based architecture. The root SKILL.md acts as a router that directs the agent to the appropriate sub-skill.

Skill Path Purpose How to Use
Router SKILL.md Entry point that routes the agent to the right review workflow. Load SKILL.md and use AgentClick UI for review.
Action Approval skills/clickui-approve/ Approve or reject risky actions before execution. Before deleting those files, show me an approval review in AgentClick UI.
Code Review skills/clickui-code/ Review shell commands, diffs, and code-related actions. Show me a code review in AgentClick UI before running that command.
Email Review skills/clickui-email/ Review inbox items, drafts, replies, and live email sessions. Open my inbox in AgentClick UI and let me triage emails.
Plan Review skills/clickui-plan/ Inspect and revise proposed plans before the agent runs them. Show me the plan in AgentClick UI before you start.
Trajectory Review skills/clickui-trajectory/ Review multi-step runs, mistakes, and resume points. Show me what you just did in AgentClick UI so I can review the steps.
Selection Review skills/clickui-selection/ Present options to the user and capture their choice. Show me the options in AgentClick UI and let me choose.
Memory Review skills/clickui-memory/ Review memory files and memory-management changes. Open memory management in AgentClick UI and let me pick which files to include.

In most cases, telling the agent to load the root skill is enough.

🛠️ Development

git clone https://github.com/agentlayer-io/AgentClick.git
cd AgentClick
npm install
npm run dev

Development mode:

  • server: http://localhost:38173
  • web: http://localhost:5173

Production-style single-port serving:

npm run build
npm start

⭐ Star History

Star History Chart

📜 License

MIT