Skip to main content
This recipe shows how to build an Agent from scratch, giving it tools (Capabilities) and the intelligence to use them (Orchestration).

​
1) Create the Agent

  1. Go to Forge > Agents.
  2. Click Create Agent.
  3. Fill in the basics:
    • Name: e.g., “Support Triage Bot”.
    • Description: What the agent does (used for routing if this agent is a sub-agent).
    • Base URL (optional): A default API host for any API Actions this agent uses.
See: Create Agent

​
2) Add Capabilities

Capabilities are the “tools” your agent can use.
  1. Open your new Agent.
  2. Go to the Capabilities tab.
  3. Click Add Capability and choose a type:
    • LLM Action: A sub-task handled by an LLM prompt (e.g., “Summarize Ticket”).
    • API Action: An external API call (e.g., “Get User Profile”).
    • Sub-Agent: Delegate to another existing agent.
See: Capabilities

​
3) Configure Orchestration (Routing)

Now define how the agent decides which Capability to use.
  1. Go to the Configuration tab → Routing sub-tab.
  2. Choose a Routing Strategy:
    • LLM Router: The model decides based on the user request and capability descriptions.
    • Deterministic: Define exact rules (e.g., “If message contains ‘refund’, call RefundCapability”).
  3. Add Orchestration Instructions to guide the LLM router (e.g., “Always check the user’s status before offering a refund”).
See: Orchestration

​
4) Define State (Agent Form)

If your agent needs to remember specific information across the session (like a user’s name or order ID), define it in the Agent Form.
  1. Go to Configuration → Agent Form.
  2. Add fields (e.g., order_id, user_email).
  3. Set Read/Write Scope to control which other agents or actions can see/update these fields.
See: Agent Form

​
5) Add Instructions

Refine how the agent plans and processes results.
  1. Go to Configuration → Instructions.
  2. Strategy Instructions: How the agent should break down complex requests (e.g., “Ask for clarification if the request is ambiguous”).
  3. Processing Instructions: How to handle results from capabilities (e.g., “Summarize the API response for the user”).
See: Special Instructions

​
6) Test in Studio

Finally, verify your agent works as expected.
  1. Go to Forge > Studio.
  2. Select your Agent.
  3. Chat with it to verify routing, state collection, and capability execution.

​
7) Call the agent from your application

Once you’re happy with the behavior in Studio, invoke the agent from code: For complex agents with sub-agents and multiple capabilities, consider passing execution flags: 
request = AgentRequest(
    agent="Support Triage Bot",
    application="YOUR_APPLICATION",
    session_id="session-123",
    execution_mode="reasoning",
    high_performance=True,
    params=ChatCompletionParams(
        messages=[{"role": "user", "content": "Handle refund for order 12345"}],
    ),
)
See Orchestration for details on execution modes and performance options.