Create Your First Agent

Once Agent Manager is installed, you can create your first agent from the console.

Choose an Agent Type

1. Navigate to Agents in the left sidebar.
2. Click Add Agent.
3. Choose how you want to get started:

• Externally-Hosted Agent — Connect an agent already running outside the platform. WSO2 Agent Manager provides observability and governance without managing the deployment.
• Platform-Hosted Agent — Deploy and manage the full agent lifecycle directly on WSO2 Agent Manager, including build, deployment, scaling, and observability.

---

Externally-Hosted Agent

Register an Externally-Hosted Agent

Fill in the Agent Details form:

Field	Description
Name	A display name for your agent (e.g., Customer Support)
Description (optional)	A short description of what this agent does

Click Register to complete the registration. The agent will appear in your agent list and start receiving observability data once you instrument it using the AMP instrumentation package.

Platform-Hosted Agent

Create a Platform-Hosted Agent

Step 1: Choose a Source Type

Select how the agent source will be provided:

• Source Code — Point to a GitHub repository containing your agent code. The platform builds and deploys it automatically.
• Agent Catalog — Pick a pre-built agent kind from the Agent Catalog.

Step 2: Fill in Agent Details

Provide a Name (e.g., Customer Support Agent) and an optional Description.

Step 3: Repository Details (Source Code only)

Field	Description
GitHub Repository	Full URL of your repository (e.g., https://github.com/username/repo)
Branch	Branch to build from (default: main)
Project Path	Path within the repository where the agent code lives (default: /)

Step 4: Build Details (Source Code only)

Choose a build type:

• Python — Provide a Start Command (default: python main.py) and select the Language Version (e.g., 3.11). Dependencies are auto-installed from requirements.txt, pyproject.toml, or package.json.
• Docker — Use a Dockerfile in the repository.

Enable auto instrumentation (Python only, checked by default)

Automatically injects OpenTelemetry tracing into your agent for observability with zero code changes. You can also pin the AMP Instrumentation Version used by the init container.

Step 5: Select an Agent Type

Choose how your agent receives requests:

Agent Type	Description
Chat Agent	Standard chat interface. Exposes POST /chat on port 8000. Request: {message: string, session_id: string, context: JSON} — Response: {response: string}
Custom API Agent	Custom HTTP API. Specify your own OpenAPI specification and port configuration.

Step 6: Configure LLM Providers (Optional)

Click + Add to bind one or more LLM providers to this agent. LLM providers are configured separately — see Register an LLM Service Provider.

Step 7: Set Environment Variables (Optional)

Add any environment variables your agent requires at runtime. Mark sensitive values as Secret to store them securely.

Click Deploy to build and deploy the agent.

---

What's Next?

Tutorial	Description
Observe Your First Agent	Set up OpenTelemetry tracing for your agent
Register an AI Gateway	Connect an LLM provider via the AI gateway
Register an LLM Service Provider	Add and configure LLM providers
Secure Agent Endpoints with API Keys	Protect endpoints with API key authentication
Evaluation Monitors	Automate quality checks for your agents
Custom Evaluators	Build and deploy your own evaluators
Configure Agent LLM Configuration	Bind LLM config to a specific agent