Agent

The Lucenia agent framework lets you automate machine learning (ML) and search tasks by combining large language models (LLMs) with reusable tools. It was extracted from the ML Commons plugin into a dedicated agent module and provides everything you need to build agentic workflows:

• Agents: Coordinators that use an LLM to reason about a problem and orchestrate the execution of tools. Lucenia supports three agent types: flow, conversational_flow, and conversational.
• Tools: Reusable building blocks that perform specific tasks, such as searching an index (SearchIndexTool), listing indexes (ListIndexTool), or running a deployed model (MLModelTool).
• Model Context Protocol (MCP): A built-in MCP server and tool registry that let external clients and agents discover and invoke Lucenia tools over stateless HTTP.

REST API namespace

Agent, tool, and MCP APIs use the preferred _plugins/_agent namespace:

• Agent management: /_plugins/_agent/agents/...
• Tool listing: /_plugins/_agent/tools/...
• MCP server and tool registry: /_plugins/_agent/mcp/...

note

The legacy _plugins/_ml/... routes for agents, tools, and MCP still work for backward compatibility but are deprecated and will be removed in a future release. Update your integrations to use the _plugins/_agent namespace.

The agent-execution endpoint (POST /_plugins/_ml/agents/<agent_id>/_execute) remains under the _plugins/_ml namespace because agent execution continues to be served by the ML Commons module.

Get started

• For an overview of agents and tools, see Agents and tools.
• For a step-by-step walkthrough, see Agents and tools tutorial.
• For the built-in tool catalog, see Tools.
• For MCP tools, see Using MCP tools.

API reference

• Agent APIs: Register, get, search for, delete, and execute agents.
• MCP server APIs: Connect to the MCP server and manage the MCP tool registry.