AG-UI (Agent User Interface) agents connect to remote HTTP endpoints that implement the AG-UI protocol, enabling integration of any AG-UI compatible server into the LLMling-Agent pool.
Overview¶
AG-UI is a protocol for building agent interfaces that provides:
- HTTP-based communication: Simple REST endpoints for agent interaction
- Streaming support: Real-time response streaming
- Standardized interface: Consistent API across different agent implementations
AG-UI agents are useful for:
- Integrating existing AG-UI compatible services
- Building distributed agent architectures
- Connecting to remote agent deployments
- Testing with locally spawned servers
Configuration Reference¶
AG-UI Agent Configuration¶
Configuration for AG-UI protocol agents.
AG-UI agents connect to remote HTTP endpoints that implement the AG-UI protocol, enabling integration of any AG-UI compatible server into the llmling-agent pool.
Example:
agui_agents:
remote_assistant:
endpoint: http://localhost:8000/agent/run
timeout: 30.0
headers:
X-API-Key: ${API_KEY}
tools:
- import_path: mymodule.my_tool_function
managed_agent:
endpoint: http://localhost:8765/agent/run
startup_command: "uv run ag-ui-server config.yml"
startup_delay: 3.0
tools:
- import_path: package.tool_one
- import_path: package.tool_two
Basic Usage¶
agui_agents:
remote_assistant:
endpoint: http://localhost:8000/agent/run
timeout: 30.0
headers:
X-API-Key: ${API_KEY}
managed_agent:
endpoint: http://localhost:8765/agent/run
startup_command: "uv run ag-ui-server config.yml"
startup_delay: 3.0
Configuration Notes¶
- The
endpointfield specifies the HTTP URL for the AG-UI agent server - Use
headersfor authentication tokens or custom routing headers - Environment variables can be used in header values:
${VAR_NAME} - When
startup_commandis provided, LLMling-Agent will spawn the server process - The server process is automatically terminated when the agent pool closes
- Use
startup_delayto give the server time to initialize before connecting - The
timeoutsetting controls the maximum time for agent responses