Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/GuaiZai233/FrostAgent/llms.txt

Use this file to discover all available pages before exploring further.

FrostAgent is an open-source Go framework that wires large language models to chat platforms. It provides an agentic execution engine with tool-call loops, persistent multi-session context, a OneBot v11 WebSocket adapter for QQ and compatible messaging platforms, and a ConnectRPC management API — all in a single deployable binary.

Quickstart

Build and run FrostAgent in under five minutes with a working LLM connection.

Configuration

Full reference for every environment variable FrostAgent supports.

Core Concepts

Understand the Engine, session management, tool registry, and adapter system.

API Reference

Explore the ConnectRPC management services and HTTP endpoints.

​
What FrostAgent does

FrostAgent sits between your LLM provider and your chat platform. Incoming messages arrive over a WebSocket connection, get routed to the agentic engine, and the engine drives a multi-step reasoning loop — calling registered tools, spawning subagents, and sending rich replies back to the platform.

Tool Registry

Register Go functions as LLM-callable tools with JSON Schema parameter definitions.

Session Management

Isolated per-user and per-group conversation history with TTL-based cleanup.

OneBot Adapter

Connect to any OneBot v11–compatible platform including QQ via WebSocket.

Vision Support

Automatic image description using a pluggable multimodal vision model.

Subagents

Delegate specialized tasks to purpose-built subagents like the Coder agent.

Management API

ConnectRPC services for monitoring bot status, sessions, logs, and settings.

​
Get up and running

1

Clone and build

git clone https://github.com/GuaiZai233/FrostAgent.git
cd FrostAgent
go build -o frostagent ./cmd/app
2

Configure your LLM provider

Copy .env.example to .env and set your upstream endpoint and API key.
cp .env.example .env
# Edit .env: set UPSTREAM_ENDPOINT and UPSTREAM_API_KEY
3

Start the service

./frostagent
# Management panel: http://localhost:8080
# OneBot WebSocket: ws://0.0.0.0:1234/ws/frostagent
4

Connect your bot

Point your OneBot-compatible client at ws://<host>:1234/ws/frostagent. FrostAgent will handle incoming messages, drive the agent loop, and send replies automatically.
FrostAgent is in early development and intended for personal and research use. Contributions and PRs are welcome.

Build docs developers (and LLMs) love

Get started for freeTalk to us