Agents

Each entry under agents: is a schema.Agent.

Minimal

yaml

agents:
  - id: assistant
    role: assistant
    brain:
      provider: deepseek
      model: deepseek-chat
      backend: openai_compat
      config:
        api_key: "{{env.DEEPSEEK_API_KEY}}"
    system_prompt: |
      You are a helpful assistant.

Fields (schema.Agent)

Field	Notes
id	Required
role	e.g. assistant, worker, coordinator, specialist
brain	Required (Brain)
max_tool_iterations	Optional cap
system_prompt / prompt	Instructions
plan_first	Optional bool
specialty	Optional label
delegate_to	Specialist ids
skills	Skill id / path string
capabilities	Per-agent capability overlay
modules	Sequence or map of module grants / tool lists
pool	max_workers, progress, auto_retry
coordination	Nested delegate_to / pool
instructions	File / specialty / capabilities helper
context	Per-agent context block
hooks	Per-agent hooks

Brain

Typical fields: provider / provider_id, model, backend, config (base_url, api_key, ...), temperature / token limits, native_tool_use, nested context, credential refs when used. Backends include OpenAI-compatible and Anthropic paths supported by the runtime LLM stack.

Roles and spawn

Coordinator agents can call agent_spawn.agent when spawn is enabled. Specialists are other agents[] entries targeted by agent / task (see Multi-agent).

Tool surface

Per-agent modules plus app tools.capabilities decide what the agent may call. Injection mode: Tools.

Related

• App config
• Multi-agent
• Behavior
• YAML building blocks