LLMling-Agent
VerifiedUnified hub for orchestrating heterogeneous AI agents through YAML configuration.
What is LLMling-Agent?
AgentPool is an open-source orchestration tool that unifies management of multiple AI agents with varying backends. It supports native PydanticAI agents alongside direct integrations and protocol-based agents, exposing them via ACP, OpenCode, or AG-UI servers.
Users configure agents, tools, knowledge sources, and workflows entirely in YAML. The system handles inter-agent delegation, parallel or sequential execution, message routing, and shared context so agents can collaborate without custom integration code.
It targets developers and teams that already work with several specialized agents and need a consistent interface for coordination, server exposure, and pipeline execution.
Capabilities
What you can build with LLMling-Agent
Sequential Code Review
Chain an analyzer, reviewer, and formatter agent to process code changes step by step and produce a polished report.
Parallel Task Execution
Run multiple coding agents simultaneously on different parts of a project and collect results in one place.
Mixed-Agent Coordination
Let a coordinator agent delegate subtasks to Claude Code, Codex, or Goose agents depending on the required capability.
Install LLMling-Agent
uv tool install agentpooluv tool install agentpool- 1Install the package with uv tool install agentpool.
- 2Create an agents.yml file defining at least one agent with model and prompt settings.
- 3Run a single agent directly from the command line using agentpool run.
- 4Start an ACP server with agentpool serve-acp agents.yml to expose agents to compatible clients.
- 5Define teams or chains in YAML and execute them from Python using the AgentPool context manager.
Works with
LLMling-Agent: pros & cons
Pros
- +Single YAML file manages agents of many different types and protocols.
- +Built-in support for delegation, parallel, and sequential workflows.
- +Multiple server protocols allow easy integration with existing tools like Zed or OpenCode.
- +Fallback models and MCP server connections are configurable without code changes.
Cons
- –Requires familiarity with YAML for all configuration.
- –Some agent types depend on external services or specific model access.
- –New agent protocols need explicit integration into the core library.
Frequently asked questions
It supports native PydanticAI agents, direct integrations such as Claude Code and Codex, ACP agents like Goose, and AG-UI agents.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar agents
Other agent frameworks options worth comparing.
