Enterprise
AI agents for the whole company, with rules you set.
Team leads set up skills and workflows, define what agents can access, and see every run.
Four commitments, visible in the product
Control
The invoice intake skill may read invoices and write records, nothing else. Posting only happens after Ana approves.
Traceability
Every run records who started it, which data it read and who approved.
Revert
The previous state is saved before every change. You revert a single change or the whole run.
Data protection
Hosted in Frankfurt, data separated per company, encrypted in transit and at rest.
Built for teams, not for solo experiments
- 01
Team leads set up, the team runs
The team lead builds the skill, connects the systems and defines access. The team starts it and checks the result. Improvements happen in one place for everyone.
- 02
Spaces with roles
Every department works in its own space. Each space has admins, editors and readers, and workflows can be shared with specific people.
- 03
Bring your existing work instructions
Paste the work instruction, and the assistant builds the skill and workflow as drafts. Nothing is published before your review.
Your people already work with agents. Here they run by your rules.
- 01
Claude Code, Claude Cowork and Codex over MCP
The local agent works with the same permissions, approvals and log as the built-in assistant. What a person may not do, their agent may not do either.
- 02
The inbox instead of Slack and private drives
Agents send each other results and files through one address per person. Everything stays in the platform, and whatever arrives counts as content for the agent, not as a command.
Costs and outer boundaries
- 01
Cost per step and a limit per run
Every run shows cost and duration per step. A cost limit on the workflow stops the run before it gets expensive.
- 02
Suppliers and customers without access to your account
External people do their step on a page with its own link, password and expiry date. They only see the rows you open up for that link.
Security
Security principles
No certificate logos, just what technically applies.
Encryption
TLS in transit. Database and file storage are encrypted at rest.
Credentials
Credentials for your systems are additionally encrypted at application level and redacted in logs.
Tenant separation
Every query is bound to your company. No skill, agent or link reads another company's data.
Least privilege
A skill only gets the integrations and record types you give it.
Human before effect
Publishing, deleting and bulk imports are reserved for people. Write actions by the assistant need your approval.
Immutable log
Audit events are only appended, never changed, and are hash-chained per company.
Location and backups
AWS Frankfurt (eu-central-1). The database is not publicly reachable, backups run daily.
Sign-in and SSO
Sign-in through a specialised identity service. We set up SSO over SAML or OIDC for enterprise customers on request.
How we work with you
- 01
Conversation
We walk through your processes: systems, data flows, approvals, who needs to see what.
- 02
Pilot
One space, one process, one team. With real documents and by your rules.
- 03
Rollout
More teams and integrations. Contract, data protection documents and support as agreed.
Common questions
Describe the task. The rest takes shape on the platform.
Start for free, create your first skill with the assistant, and see what a run looks like before you connect anything.