Tone is the open alternative to Synthflow, Vapi, and Retell. Build, deploy, and self-host voice agents that actually answer the phone — on your infra, with your models, on your terms.
Inbound · My Assistant
+1 740 520 4895 · 00:42
Built in the open · trusted by teams shipping voice AI to production
Tone collapses the voice-AI stack into one open platform — so you spend time on the conversation, not the plumbing.
Compose agents from prompts, tools, and workflows. Connect your knowledge base, set the voice, define the actions. No wrappers, no lock-in.
Push to Tone Cloud or `docker compose up` on your own infra. Same API, same runtime. Inbound, outbound, widget, or chat — your call.
Test before launch with personas. Monitor every call in production. Replay failures, tune evaluators, ship the fix — all from one place.
One platform for the whole loop: build, deploy, test, observe, improve. All open source. All replaceable.
Inbound, outbound, widget, or chat — pick the surface, write the prompt, plug in the tools. Switch LLMs (OpenAI, Anthropic, your own) without rewriting a line.
System prompt
Drop in PDFs, scrape your docs, or sync from Notion and Confluence. Tone handles chunking, embeddings, and agentic retrieval out of the box.
Customers can cancel up to 24 hours before the appointment for a full refund.
Source: Refund policy.pdf · §2.1
Visual workflow builder for multi-step calls — qualify a lead, escalate to a human, book the meeting, write to your CRM. All in one canvas.
Run thousands of simulated calls against your agent with realistic personas, accents, and edge cases. Replay real call transcripts when something breaks.
Hannah
Female · American · Professional
Chris
Male · British · Professional
Ananya
Female · Indian · Pleasant
Nick
Male · German · Angry
Voice-specific signals — interruption rate, latency, sentiment, gibberish detection — running on every call. Tune your LLM judges in Labs against ground truth.
Avg latency
218ms
-9%
Interruption rate
4.2%
-1.1pt
CSAT
4.6 / 5
+0.2
Synthflow, Vapi, and Retell are great — until you need to read the source, change a default, or move off them. Tone is what they look like when they're open. Same power, none of the lock-in.
The open-source voice AI agents platform.
Stars
9
Forks
1
Contributors
10
Commits over the last 24 weeks
# Self-host Tone in two commands$ git clone https://github.com/tonehq/tone$ cd tone && docker compose up -d✓ Tone running at http://localhost:3000✓ Workers ready · 4 lanes · model: gpt-4.1✓ Postgres · Redis · ClickHouse healthy# That's it. Point a Twilio number and ship.
No black boxes. The agent runtime, the call routing, the eval harness — all on GitHub, MIT-licensed.
docker compose up. Kubernetes Helm chart included. Your data, your network, your SOC 2 evidence.
Standard OpenAI-compatible APIs. Swap LLMs, voices, telephony providers without touching your prompts.
Roadmap, RFCs, and weekly office hours on Discord. Your PRs ship to production.
Bring your own LLM, voice, STT, telephony, and CRM. Or use the open-source connectors that ship with Tone — they're all swappable.
Don't see yours? Open an issue — most ship within a sprint.
Same code, same APIs. The only difference is who pages at 3 AM.
Free forever
Run Tone on your own infra. The full platform, zero gate-keeping.
Skip the ops. Tone-managed cloud with the same open codebase under the hood.
Custom
Dedicated infra, BAAs, on-prem deployments, custom SLAs, white-glove support.
All plans include unlimited agents and unlimited team members. Compare plans in detail → full pricing
Why we're building Tone in public, the gaps in the closed-source incumbents, and what shipping looks like from here.
How we shaved 400ms off knowledge-base lookups by making retrieval a tool the agent calls — not a wrapper around it.
Your customers don't all sound the same. The persona library that ships with Tone's test center, and how to add your own.
Spin up Tone locally in 60 seconds. Star us if it makes your day a little easier.