Live at a client
Identity and log in your tenant
Sign-in through your Entra ID, every approval in the log with the person who made it.
Claude Code in the enterprise
You have Claude Code. What is missing for it to pass IT and security.
On a developer's laptop the engine starts in a minute. In a company with real data it waits for answers to five questions about data, permissions and the trail. The answer is a layer we add on top of the engine, without replacing it. You can send this page to your IT.
In the 2025 Stack Overflow survey, 14.1% of developers use AI agents daily at work. 81% of agent users have concerns about data security and privacy. 13.8% said IT or security in their company strictly limits access to AI tools.
What Claude Code Enterprise gives you
In the Enterprise plan Anthropic adds single sign-on, automatic access removal when an employee leaves, audit logs and central settings for tool and MCP server permissions. That answers the question of who may run the agent.
We do not attack this tool. We use it and respect it. We write about the layer it does not provide, because that is not its job.
What stays open
- The agent has access to your systems: CRM, databases, file shares. Through what, and with what scope?
- The agent reads content from strangers: emails, documents, web pages. What stops it from executing a command found in an email?
- The agent can send or write something. Who approves it and where is the trail?
- Two agents work at once on the same data. Who has which scope?
- Where did the data the agent read end up?
What we add
Live at a client
Named tools instead of a console
Access to systems as a list of tools with a defined scope, read-only wherever the agent reads content from strangers.
Live at a client
A human at the end of the outbound path
Sending starts only after an employee approves. The approval is in the log.
Several agents with separate scopes
Each agent in the environment has its own permission scope. Delegation to sub-agents with verification.
Agents on employees' computers
An agent performing tasks remotely at a workstation, with the same boundaries.
The same applies to Codex and other engines. The layer is independent of the engine.
What to do tomorrow
- 011
Send your IT the security page. It answers the five questions with mechanisms.
- 022
Pick one process where the agent reads your data. Do not start with one where it sends something.
- 033
Book a demo. We show the boundaries on a running panel, with your IT in the meeting.