SIPNative, end to end
REGISTER with digest auth, re-INVITE, RFC 4028 session timers, silent drop of out-of-allowlist scanners, SIPp-tortured. It holds its own next to an SBC.
Carrier-ready voice agents
The voice agent, network-side.
A Rust daemon that plugs straight into your SIP trunks and holds the conversation: signaling, G.711 media, real-time AI — one binary, on your machines. First word in 544 ms. Licensed per channel, not per minute.
544 ms
First word (measured)
1
Binary, zero dependencies
€0
Per-minute platform fees
100 %
Of the software on your machines
The problem
Voice-agent platforms sell minutes. You run channels.
Today's SaaS platforms charge $0.07 to $0.30 per platform minute — $0.12 to $0.35 real once you add the model, synthesis and telephony — plus concurrency fees per simultaneous call. Your audio transits through their cloud, your traffic through their CPaaS, and every hop adds latency.
You don't buy your network by the minute. Why buy your voice agents that way?
The equipment
It behaves like network equipment, not like a web API.
OPSOne binary, one morning
Static Rust, embedded SQLite, TOML config, a systemd unit. No telephony stack to assemble, no middleware, no cluster to operate.
SOVSovereign, no middleman
Deployed on your premises: CDRs, recordings, transcripts and signaling never leave your infrastructure. Conversation audio goes to a single third party — the model provider you choose, under your keys and your DPA.
RTReal time, for real
Full-duplex with barge-in, first word in a measured 544 ms, RTP watchdog, answering-machine detection. The conversation doesn't lag.
OSSOperable from day one
CDRs pushed to webhooks, timestamped transcripts, WAV recordings, Prometheus-compatible /metrics and a built-in dashboard. Your tooling, not ours.
AIInterchangeable models
Real-time speech-to-speech backends (Gemini Live and compatible), brought with your own keys — billed at cost, with no hidden margin.
The economics
The platform layer disappears. What's left are your real costs, at cost.
| Per-minute SaaS platform | Strowger | |
|---|---|---|
| Billing | $0.07–0.30/min + concurrency fees | License per concurrent channel |
| One E1 at 15,000 min/month | ≈ $1,800–5,250/month all-in (0.12–0.35 $/min) | €590 license + ≈ €350 of model and telecom at cost ≈ 6¢/min all-in |
| Cost per minute | Platform $0.05–0.15 + model + telephony | Model + telephony at cost (Gemini Live ≈ $0.01–0.02), no markup from us |
| Audio and data | Transit through their platform, their CPaaS, then the model | One third party: your model provider, under your contract. CDRs, recordings and transcripts stay home |
| Telephony | Their CPaaS, marked up | Your trunks, your carrier rates |
The license is flat: every extra minute of traffic widens the gap.
Pricing
Licensed per channel. The way you've always bought capacity.
One channel = one simultaneous call, sized for your peak. The more channels you run, the less each one costs — from €69 for the first down to a full E1 (30 channels) at €590 a month.
Demo
€01 trunk · 1 channel
- 1-minute max call duration
- Your trunk, your model key
- AI email support
Per channel
€691st channel — degressive to €20
- Unlimited call duration
- Every feature, no artificial limits
- Email support
E1 · 30 channels
€590per E1 / month
- ≈ €20 per channel — the volume rate
- Every feature, no artificial limits
- Priority email support
Carrier · n × E1
Let's talkcustom terms
- n × E1 — carrier volume
- Custom volume pricing
- Priority email + emergency line
Every channel you add is cheaper than the last — the per-channel price glides from €69 to €20:
| Channels | Total / month | ≈ per channel |
|---|---|---|
| 1 | €69 | €69 |
| 2 | €118 | €59 |
| 4 | €196 | €49 |
| 8 | €312 | €39 |
| 15 | €440 | €29 |
| 30 (E1) | €590 | €20 |
Indicative pricing, excluding taxes. The license carries no per-minute cost and no concurrency fees; the per-channel price is degressive (first channel €69, ≈ €20 at a full E1). Annual billing: two months free. Model tokens and telecom minutes remain yours, paid at cost to your providers, with no markup from us — order of magnitude: Gemini Live ≈ $0.01–0.02 per conversation minute, OpenAI Realtime ≈ $0.07–0.30 depending on configuration, plus your carrier trunk rates.
Who it's for
Three ways to turn it into revenue.
TELCOCarriers
A white-label AI receptionist on every DID you sell. New ARPU on your existing base, hosted in your own core network.
CCCall centers
Overflow, night shifts, upstream qualification: one licensed channel absorbs the peak without hiring, and every conversation comes back timestamped and recorded.
SIIntegrators
An engine you can embed in your vertical solutions. You sell the usage, we supply the engine.