Skip to main content
Feather gives you a complete API to build and deploy AI-powered customer experience workflows. Create intelligent agents backed by your knowledge base, run conversations across every channel, and continuously evaluate and improve performance — all from a single, unified platform.

Quickstart

Make your first API call and run your first conversation in under 5 minutes.

Authentication

Learn how to create API keys and authenticate every request.

Build an Agent

Create your first AI agent with a knowledge base and custom tools.

API Reference

Explore all endpoints, parameters, and response shapes.

What you can build

Feather’s platform API covers the full stack of customer experience automation:

AI Agents

Create versioned agents with system prompts, knowledge bases, tools, and policies.

Knowledge Bases

Ingest documents from files, URLs, Notion, Google Drive, and S3 — then search semantically.

Multi-Channel

Handle conversations over voice, SMS, email, and API in a unified session model.

Voice Calls

Build outbound and inbound voice agents with STT/TTS, warm transfers, and voicemail detection.

Integrations

Connect Slack, Notion, Twilio, and custom MCP servers to give agents real-world tools.

Evaluations

Run simulation suites and model-judge evaluators to measure and improve agent quality.

How it works

1

Authenticate

Create an API key in your organization settings. Include it as x-api-key: <key> on every request.
2

Create an agent

POST /v1/agents with a name and system prompt. Add knowledge bases and custom tools to power its responses.
3

Start a conversation

POST /v1/v2/conversations to open a session, then POST /v1/v2/conversations/{id}/turns to send messages. Stream responses in real time with the /turns/stream endpoint.
4

Monitor and improve

Use built-in analytics, evaluators, and knowledge-gap detection to continuously improve your agents over time.

Base URL

All API requests target the sandbox environment:
https://api-sandbox.featherhq.com
Contact Feather to provision your production environment credentials. All examples in these docs use the sandbox base URL.