Architecture
An overview of BeeAI core concepts and architecture
Architecture
BeeAI enables a local-first experience, running all agents on your laptop for full data control and seamless integration with local inference engines like ollama
.
The platform is made up of several core concepts as shown in the following diagram:
Agents
An agent is a container that communicates using the Agent Communication Protocol. Agents typically define a manifest that instructs BeeAI on how to run them and may include additional features like tools.
For details on registering agents, see agents.
BeeAI includes several built-in agents, see our agent catalog.
BeeAI Server
Explore the beeai-server source code.
BeeAI server manages agents, spawns/destroys agent containers, and provides a unified ACP interface for routing requests between agents and clients.
It also manages agent configurations and environment variables and collects telemetry data which is forwarded to the Arize Phoenix OTEL backend.
The server discovers a catalog of agents from an official registry hosted on GitHub, see
agent-registry.yaml. The agents are not
downloaded automatically, a manual step beeai install <agent-name>
is required.
BeeAI CLI and UI
These components offer interfaces for interacting with agents.
Agents can have custom input/output interfaces, but the platform provides standard UIs like chat and hands-off.
Use beeai run <name>
or the graphical chat at localhost:8333 to interact with agents.
Python client
You can use the ACP SDK to programmatically interact with the agents and integrate multiple agents into a workflow for your application.
Arize Phoenix
BeeAI integrates with Arize Phoenix, an open-source agent tracing tool. For more information see agent observability.