AG2
Verifiedby AG2 AI
Builds multi-agent systems in Python with async agents, @tool functions, human-in-the-loop pauses, and a hub-and-channel Network for orchestrating agents across providers.
- Access
- API
- Auth
- None
- Skills
- 5
Skills
Define Async Agents
Creates Agent objects bound to a prompt and a model provider config, then runs a turn with ask() that returns an AgentReply.
Register Python Tools
Exposes plain Python functions to an agent with the @tool decorator so the model can invoke them during a conversation turn.
Orchestrate Agent Networks
Coordinates several agents over a hub with typed channels using the Network, the v1.0 replacement for classic GroupChat.
Pause For Human Input
Pauses a running agent to collect confirmation or missing information from a person before the workflow continues.
Compact Context And Knowledge
Layers persistent knowledge, context assembly, and history compaction onto an agent using opt-in harness primitives.
Related Agents
magentic
by Jack Collins
Turns LLM prompts into typed Python functions with decorators for structured output, function calling, and streaming…
Mirascope
by Mirascope
Calls any frontier LLM through one Python or TypeScript interface with decorators for structured output, tools, agents…
Strands Agents
by AWS
Builds production agent harnesses in Python or TypeScript with a model-driven agent loop, pluggable model providers…
Baidu Qianfan
by Baidu
Serves Baidu's ERNIE and third-party models through an OpenAI-compatible API and hosts an agent development platform…