Pricing
1 credit = €0.0001. You only pay for what runs. With your own keys (BYOK), LLM costs go through your provider; with our managed models, they are billed in credits.
Prototype your first pipeline, no commitment required.
Free
Discovery
The complete platform, with priority support and all channels.
Custom quote
Production
Dedicated deployment, SSO, audits, private models.
Custom
Regulated / sovereign
All plans include GDPR compliance, TLS/at-rest encryption, and execution logging.
Metering
All unit costs are visible per execution. Rates depend on the model and engine chosen; details are provided on request.
| Resource | Indicative rate | Note |
|---|---|---|
| LLM (input) | Per token, depending on the model | Managed models: billed in credits. With BYOK: billed directly by your provider. |
| LLM (output) | Per token, depending on the model | More expensive than input with most providers. |
| Inbound voice (ASR) | Depends on the engine | Real-time streaming (Deepgram, Whisper). |
| Outbound voice (TTS) | Depends on the engine | ElevenLabs, Cartesia, OpenAI, Deepgram Aura. Sentence-by-sentence streaming. |
| Phone call | Sum of its components | Transcription + model + speech synthesis. No per-minute price. |
| HTTP / DB operators | Included | Not metered. You pay only for LLM / voice resources. |
Definitive per-provider rates are available on request. Orchestration is not metered: HTTP operators, DB, transform, compute, code steps, and sub-pipeline calls are included. You pay only for the LLM and voice resources consumed.
FAQ
1 credit = €0.0001 (one ten-thousandth of a euro). You top up your balance in advance, and each execution debits the corresponding cost — visible to the second. No surprises at the end of the month.
Bring Your Own Key. You enter your Anthropic, OpenAI or Google Gemini API keys, or point to an Ollama, a vLLM or any OpenAI-compatible endpoint. LLM costs are then billed directly by your provider. Without a key, you use our managed models, billed in credits. BYOK remains the recommended mode in regulated environments.
On Enterprise, we support Ollama and vLLM on your infrastructure. Your prompts and completions never leave your perimeter. This is the preferred mode for regulated industries.
Executions are refused pre-call (before any debit) with an explicit message. You can configure auto-reload or budget alerts per organization.
On request, as part of the Enterprise plan. The format available today is SaaS hosted in Paris; a dedicated deployment is assessed case by case, based on your constraints.
Yes. A pipeline can call another one like a function (sub-pipelines), including across organizations with auditing. And for logic too specific for the built-in operators, a code step runs in a sandbox, bounded by a tool allowlist. These building blocks are part of orchestration and are not metered.
Yes, without disruption. Your organizations, pipelines, history, and secrets are preserved. Plan upgrades are done from the administration panel.
Atypical volumes, sovereign deployment, custom integrations — we'll build the plan that fits.
