[agent chat]
A generic AI agent service with async HTTP tool execution.
What it does
- Calls any external HTTP API (GET, POST, PUT, PATCH, DELETE) on your behalf
- Maintains session-based conversation history with token tracking
- Executes multiple tool calls concurrently in a single LLM turn
- Supports conversation compaction to stay within token limits
- Injects host secrets (API keys, tokens) transparently at execution time
- Manages API documentation for autonomous endpoint discovery
- Configurable agent bots with custom system prompts and pre-loaded docs
How it works
- Each chat request returns immediately with
state=processing
- The LLM loop runs in a background goroutine and releases when done
- Frontend polls
GET /messages?after_id=N for new messages
- Sessions resume automatically if interrupted mid-turn
JavaScript is required to use this application.