AgentGraph Documentation
The visual operating system for autonomous agents.
Orchestrate LLMs, tools, and data in a unified canvas.
Quick Start
- Navigate to the Studio via the dashboard.
- Drag an Input Node from the sidebar to the canvas.
- Connect it to an Agent Node by dragging from the handle.
- Click the Agent Node to open the Inspector and select "GPT-4".
- Click Run Workflow to execute the chain.
Core Concepts
Node Graph
A directed acyclic graph (DAG) where each node represents a computational step (LLM inference, API call, or Data transformation).
Execution Flow
Data propagates linearly. The output of Node A becomes the exact prompt/context input for Node B.
Node Reference
| Node Type | Input | Output |
|---|---|---|
| InputNode | User String | Raw Text |
| AgentNode | Context (String) | LLM Response |
| WebSearch | Query (String) | Search Results List |
Authentication
AgentGraph uses Clerk for secure identity management.
All API routes are protected by default. To consume the API externally, you must provide a Bearer Token in the Authorization header.
Note: Public access is currently disabled. Contact admin for API keys.