API-Aware Conversations
Register concise API documents once, then attach the right context to each durable chat.
Now supports any OpenAI-compatible model
Turn API Documentation Into an Agent That Gets Work Done
Give each chat its own model, API context, credentials, and operating limits. AgentChat runs the tool loop and keeps every step observable.
✓ Bring Your Own Model ✓ Durable Agent State ✓ Secrets Never Exposed to LLM
agent run / invoice-status
Processing
Context: billing APIPOST → 202 → GETState: persisted
Features
Built for long-running, real-world API work
Built for developers and automation enthusiasts who want AI to do the heavy lifting.
Asynchronous Polling
Agents can wait, honor retry guidance, poll long-running jobs, and respond immediately when you stop them.
Per-Chat Header Injection
Attach host-specific headers to each chat. Matching HTTP requests receive them without exposing values to the LLM.
Durable, Observable Runs
Streaming text, token usage, tool calls, and tool results are persisted so clients can follow every step.
Workflow
Up and Running in 3 Steps
Connect Your Model
Save one or more OpenAI-compatible AI services and choose one for the chat.
Set the Operating Context
Attach API documents, host headers, a system prompt, and execution limits.
Run and Observe
Describe the outcome. AgentChat streams and persists every model and tool step.
Developer API
Embed the agent loop without rebuilding it
Create chats with your API key, then query durable state and incremental messages from your own product. AgentChat handles execution, recovery, and concurrency for you.
Open the developer dashboard →https://agent.iamdev.me
POST/api/agent/sessions
POST/api/agent/sessions/{id}/chat
GET/api/agent/sessions/{id}/messages?after_id=...
POST/api/agent/sessions/{id}/stop
AgentChat