Quick Start
Get Karate Agent running in under 60 seconds.
Prerequisites
- Java 21+ installed
- Docker running (Docker Desktop on macOS/Windows, or Docker Engine on Linux)
1. Pull the Worker Image
docker pull karatelabs/karate-agent:latest
2. Start the Grid Server
java -jar veriquant.jar grid --port 4444
Open http://localhost:4444 in your browser — you’ll see the Karate Agent dashboard.
3. Start an Interactive Session
From the dashboard, click New Session to start a browser container. Or via curl:
curl -X POST http://localhost:4444/api/sessions \
-H "Content-Type: application/json" \
-d '{"mode": "interactive"}'
The response includes a sessionId. Use it to send commands:
# Navigate to a page
curl -X POST http://localhost:4444/sessions/{sessionId}/proxy \
-d '{ "js": "agent.go(\"https://example.com\")" }'
# Discover elements
curl -X POST http://localhost:4444/sessions/{sessionId}/proxy \
-d '{ "js": "agent.look()" }'
# Click a link
curl -X POST http://localhost:4444/sessions/{sessionId}/proxy \
-d '{ "js": "agent.act(\"{a}More information\", \"click\")" }'
4. Watch Live
In the dashboard, click the session to see the live browser view via noVNC. You can watch the browser in real-time as commands execute.
5. Connect Your LLM Agent
Via MCP (Claude Code)
claude mcp add karate http://localhost:4444/mcp
Via REST (any agent)
The /sessions/{id}/prompt endpoint serves a self-contained API reference your LLM can read. Point your agent at it:
curl http://localhost:4444/sessions/{sessionId}/prompt
Next Steps
- Interactive Mode — Deep dive into interactive exploration
- Flows — Create reusable automation scripts
- Autonomous Mode — Submit jobs for unattended execution