Overview
Each endpoint on ACN is a specific action your service can perform. Agents discover endpoints by their description and call them using their parameter schema. Well-defined endpoints get better discovery matches and more usage.Endpoint Structure
Every endpoint is defined with:Writing Good Descriptions
The description is what the discovery engine uses to match agents to your endpoint. A good description:Be specific about what the endpoint does
Be specific about what the endpoint does
Good: “Send a transactional email with HTML support and delivery tracking”Bad: “Email endpoint”
Include key capabilities
Include key capabilities
Good: “Get real-time cryptocurrency prices in any fiat or crypto denomination, supports 10,000+ coins”Bad: “Get prices”
Mention limitations or constraints
Mention limitations or constraints
Good: “Translate text between 100+ languages, max 5,000 characters per request”Bad: “Translation service”
Parameter Best Practices
Clear Descriptions
Every parameter should have a description that an LLM can understand:Use Standard Types
| Type | When to Use |
|---|---|
string | Text, IDs, URLs, emails, enums |
number | Integers, floats, coordinates |
boolean | On/off flags |
object | Nested structures |
array | Lists of items |
Required vs Optional
Mark parameters asrequired: true only if the endpoint cannot function without them. Optional parameters with sensible defaults make the endpoint more flexible.
Response Schema
Define your response schema so agents know what to expect:Choosing Tags
Tags help with structured filtering. Use lowercase, specific terms:- Good tags:
email,transactional,smtp,html-email - Bad tags:
api,service,fast,best
Multiple Endpoints
If your service offers multiple capabilities, create separate endpoints for each:| Instead of… | Create… |
|---|---|
One “email” endpoint with a type parameter | send-email, send-batch, check-status |
One “data” endpoint with an action parameter | get-price, get-historical, get-market-cap |
Testing Your Endpoints
Before going live, your endpoints are tested on ACN’s testnet:- Discovery matching — Does your description surface for relevant queries?
- Schema validation — Do requests validate correctly?
- Error handling — Do errors return clear, structured messages?
- Response format — Does the response match your schema?