Lyra API
The same private pipeline behind the Lyra app, as an API: it routes each request to the cheapest capable model, redacts personal data before any cloud call, and returns a portable, signed audit record of what happened.
What the API gives you
- Smart routing — each request is classified and sent to the cheapest capable model (local or cloud); you see which model answered.
- PII redaction — personal identifiers are detected and redacted (Presidio) before any cloud call. Best-effort, not a guarantee — see the Trust Centre.
- Bring your own key — thread your own provider keys; we don't store your API traffic.
- Signed audit record (Trust Envelope) Planned — a tamper-evident record of every request (model, cost, redactions), portable across your tools.
Authentication
Bearer token per request:
Authorization: Bearer <your-api-key>
OpenAI-compatible endpoint, so existing SDKs work with the base URL swapped. The base URL and a copy-paste quickstart are shared when your API access is approved.
Pricing (per token, AUD)
Transparent per-token rates. Local model runs are unmetered.
| Model | $/M tokens in | $/M tokens out |
|---|---|---|
| Lyra Local (on-device) | Unmetered | Unmetered |
| Claude (Sonnet-class) | $7.60 | $38.00 |
| GPT (frontier) | $6.35 | $25.35 |
| Gemini (Pro) | $3.20 | $12.70 |
| Perplexity (search) | $2.55 | $2.55 + search fee |
All prices in AUD, excl. GST. Third-party model prices compared as at 4 September 2026; providers may change their rates at any time. A routing, PII-redaction and audit fee of ~40% applies on top of the underlying model cost. Bring your own key and it's pass-through — no markup.
Request API access →