Build

Build Your Own Agent

A complete guide to building, testing, and deploying intelligent agents on BeeSystem open infrastructure. From spec to production.

01

Choose a capability

Start by defining what your agent should be able to do. What domain does it operate in? What tasks will it handle? What outputs should it produce?

Example: A research agent that reads documents, summarizes content, and produces structured reports.

02

Define agent behavior

Write your agent specification. The spec defines the agent's capabilities, behavioral constraints, and output format in an open, auditable YAML document.

name: my-research-agent
version: "0.1.0"
capabilities:
  - reason
  - use_tools
  - remember_context
constraints:
  - no_silent_fallbacks
  - structured_errors_only
03

Connect tools

Wire your agent to the tools it needs. BeeSystem includes a built-in tool library. Use the Bee SDK to build custom connectors for any API or service.

Available tools: web_search, document_reader, structured_writer, api_connector, database_query, and more via SDK.

04

Add memory and context

Configure how your agent manages state. Choose session memory for stateless tasks, persistent memory for stateful workflows, or retrieval-augmented patterns for knowledge-heavy agents.

Configure memory type, storage backend, namespace, and retention policy in your agent spec.

05

Test locally

Use the Bee Worker runtime locally to test your agent before deployment. The local runtime is identical to production — no environment drift.

bee worker start
bee agent run \
  --spec ./my-agent/agent.yaml \
  --task "Summarize this document"
06

Deploy with Bee Worker

Deploy your agent using Bee Worker runtime containers. Works with Docker, Kubernetes, and any container platform. One spec, any environment.

docker run \
  -v ./agents:/agents \
  beesystem/bee-worker:latest
07

Scale with Bee Hive

For production workloads, add Bee Hive for distributed orchestration. Route tasks, manage agent pools, and coordinate multi-agent workflows at scale.

Bee Hive handles routing, load balancing, and agent pool management automatically.