Hail documentation

Hail gives your AI agent a voice, a real phone number, and an inbox — place AI phone calls with ElevenLabs voices, run an AI call center, send SMS and agent mail, all behind one MCP endpoint, one API key, one invoice. Most people use Hail Cloud; Hail's services are also self-hostable under AGPLv3, with LiveKit Cloud and channel providers remaining external.

Using Hail Cloud

• MCP clients — connect Claude.ai, ChatGPT, Cursor, or any MCP client. Paste a URL, click Allow, done.
• Agents — save an agent once; numbers answer calls and texts with it, and POST /v1/calls can place calls with it.
• Webhooks — signed JSON events for inbound mail, SMS, delivery reports, and call outcomes.
• Email delivery: what Hail does so mail reaches the inbox: signed domains, one-click unsubscribe, bounce and complaint suppression, events and stats.
• react-email — render a react-email template and send it through POST /v1/emails.
• Bring your own LLM — point voice calls at your own OpenAI-compatible endpoint.
• CLI reference — the hail binary's email and webhooks surface.
• API reference — every REST endpoint, generated from the OpenAPI spec.

Understanding Hail

• Architecture — the three Python services, the Go CLI, and how LiveKit Cloud fits in.

Running it yourself

• Self-hosting — the umbrella: one-VM deploy, provider setup (LiveKit, Twilio, SES), SMTP inbound, and the operations runbook.
• Contributing — dev environment, regenerating OpenAPI, PR flow.

Source lives at github.com/hail-hq/hail. Every page here is also plain markdown in docs/public/ — anything in that folder is published, anything outside it is not.

• Optional call recording: settings, playback, billing and self-hosted storage.