Your API already works.Give it an agent.

1version: "2"
2assistant:
3  language: auto
4models:
5  routing:
6    provider: openai
7    model: YOUR_TOOL_CAPABLE_MODEL
8    api_key: $ENV:OPENAI_API_KEY
9tools:
10  - id: find-orders
11    tool:
12      description: Find the current user's orders by product name.
13      input_schema:
14        type: object
15        properties:
16          query: { type: string, minLength: 1 }
17        required: [query]
18    request:
19      method: GET
20      url: https://api.example.com/orders
21      auth: { type: session }
22      map:
23        query:
24          q: $.query

Replace the example URL and model placeholder. Set OPENAI_API_KEY in your backend environment. The search uses the current user’s API token.

1import { AgentRuntime } from "agento-runtime";
2 
3// Create once when your backend starts.
4const runtime = await AgentRuntime.create({
5  configPath: "./config.yml",
6  presentation: "raw",
7});
8 
9// The user message to the agent.
10const clientMessage = "Find my headphone order";
11// AGENTO selects find-orders and extracts its inputs.
12const result = await runtime.chat({
13  sessionId: `user-${user.id}`,
14  auth: {
15    token: user.accessToken,
16  },
17  message: clientMessage,
18});
19 
20console.log(result);

user comes from your authenticated backend. chat() uses the routing model in config.yml to select the tool; your API enforces the user’s permissions.