Pricing

One plan. It’s free.

The fuze-aiSDK is open source under MIT — free, today and forever. The hosted dashboard is free too while we’re in early access: no card, no commitment. That’s the whole offering right now — one plan, no tiers. When we introduce paid plans you’ll get at least 60 days’ notice, and if you need higher limits before then, get in contact.

Free

Early access
€0no card required

Everything Fuze ships today, on one plan. The SDK is MIT and free forever; the hosted dashboard is free while we’re in early access.

  • The full fuze-ai SDK — guard(), loop(), budget(), side_effect(), requestOversight()
  • Append-only JSONL evidence stream — Art. 12 event recording
  • HMAC-SHA256 hash chain in the Python SDK; TypeScript parity on the roadmap
  • Token usage extraction across OpenAI, Anthropic, Mistral, Vercel AI SDK, Mastra, LangChain
  • Python framework adapters for LangGraph and CrewAI
  • Hosted dashboard in Frankfurt — runs, agent health, retention, admin audit log (Art. 26)
  • Annex IV bundle, FRIA draft, and Article 73 incident generators
  • EU data residency — Frankfurt. Evidence stays independently verifiable with the open-source verifier

Need more?

on request
Get in contact

We’re free during early access with generous limits. If you’re running at higher volume or need guarantees in writing, reach out and we’ll work something out.

  • Higher trace limits and volume
  • EU data-residency guarantee, SLA, and SSO
  • Negotiated DPA / Master Services Agreement
  • On-prem deployment and sovereign provider routing — Mistral, Scaleway, OVHcloud
  • KMS-backed Ed25519 signing and custom Article hooks
Get in contact hello@fuze-ai.tech

What’s included

What you get on the free plan.

Grouped by where each capability lives — in the open-source SDK, in the hosted dashboard that’s free during early access, or available on request. No checkmark matrix; if it’s not in the paragraph below, it isn’t shipping.

Free · open source

The open-source primitives

The fuze-ai SDK for TypeScript and Python: guard(), loop(), budget(), side_effect(), and requestOversight(). Every prompt, tool call, return value, and guard event is written to an append-only JSONL evidence stream. The Python SDK adds an HMAC-SHA256 hash chain on top; TypeScript parity is on the roadmap. The SDK wraps raw OpenAI, Anthropic, and Mistral calls today; Python ships LangGraph and CrewAI adapters. MIT-licensed. Self-host or use the hosted dashboard on top of it — your choice.

Free · hosted

Hosted backbone and compliance packages

A managed evidence backbone in Frankfurt that reads the SDK’s stream and turns it into the artefacts regulators ask for — free while we’re in early access. @fuze-ai/agent-annex-iv maps evidence span attributes to Annex IV sections and ISO 42001 controls. @fuze-ai/agent-fria compiles a Fundamental Rights Impact Assessment from the same stream. @fuze-ai/agent-incident drafts a regulator-ready Article 73 incident report. The dashboard surfaces runs, agent health, retention windows, and an admin audit log of privileged actions (Art. 26).

On request

Sovereign deployment and procurement support

Not part of the free plan, but available if you need it — just get in contact. On-prem deployment via @fuze-ai/agent-sovereign-terraform, sovereign provider adapters for Mistral, Scaleway, and OVHcloud, KMS-backed Ed25519 signing, and custom Article hooks or policy adapters (Cerbos integration available). DPA, MSA, and procurement support handled directly — useful if your legal team needs negotiated terms before the August 2026 enforcement date.

What Fuze does and does not do

Who is responsible for what.

Fuze provides tooling and evidence. We are not a regulator, an auditor, or a notified body. Using Fuze does not by itself make any AI system compliant with the EU AI Act, GDPR, or any other regulation. Under Art. 25(4) we are a component supplier.

Under the EU AI Act, the deployer (and the provider, if you place a system on the market) holds the regulatory obligations. Under the GDPR, the controller determines purposes and means of processing. In all of these roles, you are responsible for your compliance posture. Fuze produces the evidence and workflows that help you meet it.

The Services are provided “as is” without warranty. Aggregate liability is capped at fees paid in the preceding twelve months, which while we’re free is zero. See Terms of Service §10–§12 and the deployer-vs-provider guide.

Nothing on this page or in the Services constitutes legal advice. Engage qualified EU counsel for advice on your specific compliance obligations.

FAQ

Questions teams ask before signing up.

Q.

Is it actually free, or is this a trial?

A.

Genuinely free. The fuze-ai SDK and the surrounding @fuze-ai/* packages are MIT-licensed — read the source, fork it, run it locally, emit evidence to your own store. The hosted dashboard is free too while we’re in early access. No time limit and no feature lockouts; one plan, and it’s the free one.

Q.

Do you have paid plans?

A.

Not yet. Right now there’s a single plan and it’s free — the SDK forever, the hosted dashboard during early access. When we do introduce paid plans we’ll give at least 60 days’ notice before anything changes. If you need higher limits or guarantees in writing in the meantime, get in contact.

Q.

Does using Fuze make my AI system EU AI Act compliant?

A.

No. Fuze provides tooling and evidence. The deployer (and provider, if you place a system on the market) is responsible for compliance. Fuze produces signed, hash-chained evidence that your obligations are being met, but the obligations remain yours. See Terms of Service §10 and the deployer-vs-provider guide.

Q.

Where does the data live?

A.

The open-source SDK runs in your infrastructure — evidence emits wherever you configure, and no data leaves your environment by default. The hosted dashboard is in Frankfurt. If you need to run the backbone on-prem via @fuze-ai/agent-sovereign-terraform or route inference through sovereign providers (Mistral, Scaleway, OVHcloud), get in contact.

Q.

Can I take my evidence with me?

A.

Yes. Evidence is hash-chained and Ed25519-signed, and the audit log format is a published spec. If you stop using the hosted dashboard, the evidence you generated remains independently verifiable with the open-source verifier.

Q.

We have a Master Services Agreement template. Will you sign it?

A.

Probably yes. We handle DPA, MSA, and procurement directly — email hello@fuze-ai.tech with the template and we’ll come back.

Next

Start with the classifier, or get in contact.

If you’re not sure whether the Act applies to your agent, start with the risk classifier. If you already know it does and you need higher limits or guarantees in writing, reach out.