Agentic voice

Voice agents, built and run by AI agents

Describe the agent in a file. Your coding agent validates it, compiles it, deploys it and reads back what happened on the call.

Built with the leading voice ecosystem

How it works

From a file to a live call, without opening a dashboard

UNMUTE describes the agent. SLNG runs it. Your coding agent writes the file, validates it against your rules, compiles it and deploys it through the API, with SLNG routing every call underneath in the region you chose.

Define

One UNMUTE file holds the whole agent. Prompt, models, tools, channels. Your coding agent writes it, you review it like any other code.

Govern

A saved manifest sets your company's rules: which models and providers, which languages and regions, which deployment targets and tools.

Compile and ship

Violations fail at compile, not on a call. The same package targets Pipecat, LiveKit or SLNG, then deploys, versions and rolls back through the API.

Agent capabilities

What an AI agent can do today

Your agent can preview a change, roll it back, and read what happened on a real call.

Public OpenAPI specs

Reads the Voice Agents and Management APIs directly without the guesswork.

Versioned voice agents

Lists versions, previews a restore before applying it, accepts it, labels versions, downloads a version’s config.

Preview tool upgrades

Previews an upgrade to see the parameter diff and blocking issues, then accepts with the returned hashes.

Calls API

Dispatches outbound calls with status webhooks, fetches a call, exports calls.

Session detail

Reads which tools the voice agent called and what each returned.

Region and runtime

Chooses where a voice agent runs and moves it for future calls.

Custom code tools

Writes typed Python tools that run in an isolated sandbox.

MCP tools for voice agents

Connects an MCP server so the voice agent can use its tools during calls.

Secrets vault

Stores credentials the voice agent needs; secret values are never returned.

Model catalog

Lists speech models by capability, provider, hosting and language.

Usage

Tracks spend, audio minutes and requests.

ISO 27001

Certified

SOC 2 Type II

Certified

HIPAA

Compliant

GDPR

Compliant

In-region compute

Your agent picks the region

Region is a field in the package and a parameter in the API. Your coding agent sets it at deploy and moves it for future calls, without anyone opening a dashboard.

Base URL

https://eu-west.api.slng.ai

15 regions, local compute.

Get started

Hand this to your coding agent. Change the region to wherever your callers are.

Prompt

Build a voice agent with UNMUTE and deploy it on SLNG. Install the UNMUTE CLI and skill, create a package from our manifest, run unmute validate and unmute compile, then deploy to SLNG in region eu-west. Docs index: https://docs.slng.ai/llms.txt. UNMUTE: https://unmute.ai.

Can my coding agent build a voice agent end to end?

Yes. With the UNMUTE CLI and skill it describes, validates and compiles the voice agent, then deploys and operates it through the SLNG API.

Do I have to run on SLNG?

No. UNMUTE compiles to Pipecat and LiveKit too, and SLNG can serve the models underneath either.

How do I keep AI agents within our rules?

Save a manifest with the models, providers, languages, regions and tools you allow. Validation fails on anything outside it.

Is UNMUTE open source?

Yes, and it works fully locally for free.