Speko Docs

Speko developer platform

Build and operate production voice AI.

Run managed agents across browser and telephony, then choose the routing, runtime, SDKs, and agent tooling that fit your architecture.

Managed Agents Platform

From persona to production call

Create reusable voice agents with their own prompt, voice, routing preferences, tools, and knowledge. Deploy them once, then use the same agent across browser sessions and phone calls.

Define the persona and stack
Persist the system prompt, voice, language, model settings, and STT, LLM, and TTS provider preferences.
Connect tools and knowledge
Give each agent its own registered tools and knowledge, including managed search, call transfer, and hangup actions.
Run across browser and phone
Use the same agent in browser conversations and inbound or outbound phone calls, with per-session overrides when needed.
Operate the full lifecycle
Deploy versions, test sessions, and inspect transcripts, recordings, lifecycle events, reports, and costs from one control plane.

Choose your integration boundary

Each surface solves a different layer of the stack. Start with one and combine them as your product grows.

How the pieces fit

The Agents Platform runs the product layer across browser and phone. Relay and Gateway cover execution paths, SDKs integrate them, and MCP lets coding agents operate the control plane.

Managed agents

Speko runs the application layer.

Browser or phone

Speko Agent

Tools and knowledge

Hosted

Speko runs the routing layer.

Your app or SDK

Speko Relay

Voice provider

Provider direct

You run the routing boundary.

Your voice agent

Speko Gateway

Voice provider

Agentic

Your coding agent operates Speko.

Codex or Claude

Speko MCP

Speko control plane