No ceilings
Agents, flows, executions, integrations and scheduled work are all uncapped. The platform stops being something you ration.
Autonomy your auditors can follow.
Agents that execute across your stack, with a gate on every irreversible action and a step-by-step record of everything they did. Scoped, contracted and priced per engagement.
Fleece AI for Enterprise is a negotiated engagement rather than a plan you buy with a card: unlimited agents, flows and executions, 365 days of execution history, an append-only audit log, approval gates on every irreversible action, and every frontier model including Claude Fable 5. Deployment, terms and support are scoped in the contract.
This is the right conversation if
- You need agents that act on production systems, and a record of every action they took.
- A human must approve anything irreversible — payments, customer communication, publishing.
- Your security review needs specifics: retention, residency, access, and who can do what.
- You want terms, deployment and support written down rather than accepted from a checkout page.
- The work spans many systems, and a single-vendor assistant would only cover part of it.
What the product gives you
Approval gates per action
You decide which actions require a human yes. The agent pauses at that gate and waits — it does not proceed and report afterwards.
An append-only audit trail
Security-relevant actions are written to an append-only log, and every run keeps a step-by-step trace of what the agent read, decided and wrote.
365 days of execution history
A full year of run history retained, so an investigation months later still has something to read.
SSO, including SAML
Sign-in through Google or Microsoft Entra ID today, and SAML for your own identity provider as part of the engagement.
Every frontier model
Including Claude Fable 5, plus EU-processed models where data residency matters to your review.
Priced per engagement
Enterprise is quote-based. What it includes — deployment, terms, support and the identity integration you need — is scoped with you rather than picked from a page. For reference, Business is the highest self-serve plan.
Questions a security review asks
- What exactly is recorded when an agent acts?
- Every run keeps a step-by-step trace: the tools called, the arguments, the results and the decisions between them. Separately, security-relevant actions are written to an append-only audit log. Together they answer 'what did it do, and why' months after the fact.
- Can we stop an agent before it does something irreversible?
- That is the default posture. You choose which actions require approval — moving money, contacting a customer, publishing — and the agent holds at that gate until a person clicks. Nothing proceeds on the assumption that you would have said yes.
- Where is our data processed?
- Data is encrypted in transit and at rest and is never used to train models. Model processing depends on which model an agent uses; EU-processed options are available where residency matters. The specifics for your deployment are part of the engagement.
- What is your certification status?
- SOC 2 Type II is in progress, and we say so rather than implying otherwise. GDPR, CCPA and EU AI Act compliance are in place. If a certificate is a hard gate for your procurement today, tell us early — it is a fair question and it changes the conversation.
- How long do you keep execution history?
- 365 days on Enterprise, purged automatically after that by a scheduled job. Custom retention is part of what gets scoped in the contract.
- Can agents be limited to specific systems?
- Yes. Agents only reach what you connect, and every connection is an explicit OAuth grant you can revoke. There is no ambient access to systems nobody wired up.
Start with the security review, not the demo
Tell us what your auditors need to see and what the work actually is. We will scope the deployment, the terms and the identity integration around that.
Talk to us