For the complete documentation index, see llms.txt.
Skip to main content
Version: 8.10

Build AI agents

Build AI agents in Camunda by modeling a Camunda AI agent or connecting an external agent.

About​

Ready to start building AI agents in Camunda? Start by choosing what type of agent you want to build:

Agent typeDescription
Build a Camunda AI agentModel the agent in BPMN, connect it to an LLM, and give it tools and memory.
Connect an external agentOrchestrate an agent that runs outside Camunda, such as in LangGraph or on Amazon Bedrock.

Build a Camunda AI agent​

A Camunda AI agent runs its agent loop in Camunda's engine, which activates each tool call as a governed BPMN activity. You model the agent with the AI Agent connector.

tip

Choose an LLM​

A large language model (LLM) provides the agent's reasoning. The agent sends it a prompt and the available tools, and the LLM decides what to do next.

Extend your agent capabilities​

Use these guides to add tools and long-term memory to your agent.

Tools are the BPMN activities an agent can call to act on your systems, such as service tasks, human tasks, and sub-processes. The agent selects which tools to run, and Camunda applies retries and incident handling to each call.

By default, an agent only knows what is in its conversation context. To give it long-term memory, store embeddings in a vector database and retrieve them when the agent needs them.

Connect an external agent​

An external agent runs its agent loop in an external runtime instead of Camunda's engine. Camunda orchestrates when the agent runs as part of your process, and you connect it so its activity is visible in Camunda.

Expose a process to other agents​

Agents built in Camunda or elsewhere can also call your processes. With the Model Context Protocol (MCP), you can expose a BPMN process as a tool that agents discover and run.

Build your first AI agent
Get started with Camunda agentic orchestration by building and running your first AI agent.
Get started