Skip to main content
Feather helps you build AI-driven customer conversations across voice and messaging channels from one platform. A single agent model powers live calls, SMS threads, email conversations, web chat, workflow-driven outreach, and handoffs between specialized agents.

What You Can Build

Voice Agents

Handle inbound calls, launch outbound calls, transfer conversations, and capture call outcomes.

Messaging Agents

Run SMS, email, and web chat conversations with the same prompt, tools, and variables model.

Automated Outreach

Orchestrate multi-step call and text workflows with scheduling, throttling, and execution controls.

Booking and Routing

Route calls through squads, connect calendars, and let agents schedule appointments in flow.

Core Building Blocks

Agents and Versions

An agent is the long-lived identity for a use case such as support, sales, intake, or reminders. An agent version is the deployable configuration for that agent: prompt, tools, model settings, voice, language, and channel behavior.

Conversations

Feather supports four main conversation surfaces:
  • Voice calls
  • SMS threads
  • Email threads
  • Web chat threads
Each surface can use shared variables, metadata, tools, and knowledge sources.

Workflows

Workflows let you automate outreach over time. A workflow defines the sequence. Each execution runs that sequence for one lead or contact.

Tools and Knowledge

Tools let an agent read or write data in external systems. Knowledge base collections let an agent ground answers in your documents and internal content.
  1. Create an agent for the job you want to automate.
  2. Add tools, prompt variables, and optional knowledge base collections.
  3. Configure phone numbers, email sending, chat domains, or calendars for the channel you need.
  4. Deploy the agent version you want to expose to live traffic.
  5. Test the behavior in Testing Lab before scaling through workflows or public entry points.

Start Here

API Reference

When you are ready for endpoint-level detail, start with: