You’ve built an autonomous agent. It works… until it doesn’t.
- Infinite retry loops drain your API budget while you sleep
- Hallucinated steps cause agents to click the same button 47 times
- Timeout failures leave you parsing 5,000 lines of JSON logs at 2 AM
- No visibility into what went wrong or where the agent got stuck
Empusa fixes this.
Real-time overview of all running agents with status indicators and intervention counts
Step-by-step execution timeline with success/failure states and loop detection
Visualize execution traces as an interactive timeline. Green = Success, Red = Fail, Yellow = Loop Detected. No more grep-ing through logs.
The system automatically identifies when an agent is stuck in a retry loop and shows exactly where intervention occurred. Watch the yellow pulse in action.
Pause bad runs, inspect the state, fix the issue, and resume without restarting from scratch. Save hours of debugging time.
Built for OpenClaw logs but designed to work with any Agent Protocol (LangChain, AutoGPT, custom frameworks). Plug in your telemetry and go.
This tool was born from fixing the critical Infinite Loop bug in OpenClaw (PR #9759). After watching agents burn through $200 in API credits overnight, we realized the ecosystem needed better observability.
We built this because we needed it. Now we’re open-sourcing it for the community.
# Clone the repository
git clone https://github.com/justin55afdfdsf5ds45f4ds5f45ds4/EmpusaAI.git
cd EmpusaAI
# Install dependencies
npm install
# Run the development server
npm run dev
Open http://localhost:3000 to see the dashboard.
Run the test script to send sample logs to the dashboard:
node scripts/test-ingest.js
This will create a test session with 5 log entries including a loop detection. The script will output a URL to view the session.
Watch a real-time agent simulation with dynamic state evolution:
node scripts/simulation-agent.js
This runs a live “Flaky Browser Agent” that:
- ✅ Navigates and clicks successfully
- 🔄 Retries a failing action 3 times
- 🚨 Detects the loop and halts automatically
- 💾 Updates state in real-time (watch retries increment: 0 → 1 → 2 → 3)
Open the dashboard while it runs to see logs appear live!
Test the resume capability:
# Copy the example resume file
cp resume.example.json resume.json
# Run the simulation - it will skip completed steps!
node scripts/simulation-agent.js
The agent will:
- ♻️ Load state from
resume.json - ⏩ Skip Steps 1 & 2 (already completed)
- 🔄 Resume from Step 3 onwards
This proves you can restore agent state and continue execution!
Ingest agent execution logs in real-time.
Request Body:
{
"sessionId": "session-123",
"step": 1,
"action": "Click(Login)",
"status": "success",
"error": "Optional error message",
"timestamp": "2026-02-06T10:42:01.000Z"
}
Status Values:
success– Step completed successfullyfailure– Step failed with errorloop_detected– System detected infinite loop
Response:
{
"success": true,
"id": 1
}
Retrieve all logs for a specific session.
Response:
{
"logs": [
{
"id": 1,
"session_id": "session-123",
"step": 1,
"action": "Click(Login)",
"status": "success",
"timestamp": "2026-02-06T10:42:01.000Z"
}
]
}
| Technology | Purpose |
|---|---|
| Next.js 14 | React framework with App Router |
| TypeScript | Type-safe development |
| Tailwind CSS | Utility-first styling |
| Lucide Icons | Beautiful, consistent iconography |
We welcome contributions! Whether it’s:
- 🐛 Bug reports
- 💡 Feature requests
- 📖 Documentation improvements
- 🔧 Code contributions
Check out our Contributing Guide to get started.
MIT License
Copyright (c) 2026 EmpusaAI
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
See LICENSE file for full details.
Built by developers who got tired of debugging agents in production. Special thanks to the OpenClaw community for inspiration and feedback.


