One route across
your voice providers.

Route STT, LLM and TTS through one API. Speko measures the options, applies your policy and falls back only where recovery is safe.

See the route contract Run a paid pilot

One contract for the full stack.

English / balanced / cascade

Speko filters supported models against your requirements, then ranks eligible candidates using measured data.

Compare providers
  1. 01
    Speech to text

    Streaming, language, latency and accuracy.

  2. 02
    Language model

    Tool support, quality and first-token latency.

  3. 03
    Text to speech

    Voice, streaming, region and first-audio latency.

Plug in. Keep your stack.

LiveKit

Use the shipped adapter in your existing AgentSession.

pnpm add @spekoai/sdk @spekoai/adapter-livekit
Open LiveKit quickstart

Pipecat

Keep Pipecat and map custom processors to Speko's HTTP APIs.

STT  POST /v1/transcribe
LLM  POST /v1/complete
TTS  POST /v1/synthesize
Open API docs

Keep your stack.
Give Speko a measured slice.

Start with one workload and one bounded traffic ramp.

Design the pilot Read the gateway guide