Skip to content

Only the people you choose use the tool,and your AI never sees your credentials.

A tool limited to your colleagues shows a stranger nothing but its sign-in page, and nothing important happens without your approval.

A conversation with your AI
  1. You: Share the tool with the team, and connect it to our payment software.
  2. Your AI: Done. Enter the payment credential in while. I will never see it.
  3. You: Delete the old tool’s data.
  4. Your AI: Requested. The deletion is waiting for your approval. Delete the old tool’s datarequested by your AI to approve
Works with

The message is the same for every AI.

Hosted in Franceby Nextmap

Everything is in place from day one

Your credentials stay out of the AI’s reach

You enter another piece of software’s credential in while. The tool uses it without ever reading it, and your AI never sees it.

  • Creating, changing or deleting anything in your other software is blocked until a person has allowed it.
  • For the first 7 days, the tool’s calls are recorded. After that, only calls of the same kind go through.
  • A leaked credential is replaced without changing a line of code.

You decide what waits for your approval

By default, your AI updates the tool without waiting for you. What matters always needs a person’s approval, and the signed history keeps everything.

  • Deleting or restoring production data.
  • Raising a budget or changing a rule.
  • A call you marked “approve every call”, like a refund.

Each tool runs apart, and your AI is warned about its known vulnerabilities

When a serious vulnerability is known, your AI can update the dependency.

  • A tool can’t reach the tools of another project.
  • Each version’s dependencies are listed and checked against known vulnerabilities at every new version, then every day.

You choose who uses each tool

Everyone, a client with a link and a password, or your colleagues only.

  • On the Pro and Enterprise plans, your colleagues sign in with their work email account.
  • A link for a client lasts 1, 7 or 30 days, with a password.

You go back, in the tool and in the data

Your AI goes back to the previous version, and your data goes back to a specific point in time.

  • Data is backed up every day, and in production, one backup a day is checked by a real restore.
  • Backups are encrypted with AES-256 and kept for 14 days.

Your data stays in France, with a French company

Your tools and their backups are in France, with Nextmap.

For your security questionnaire

Builds
Each build runs in its own small virtual machine.
Sign-in and network
The company sign-in goes through OpenID Connect, with your company’s provider. Everything goes over HTTPS, with HSTS by default. Over the internal network, a tool can’t reach any tool in another project, and its outbound traffic to the internet is blocked by default.
Secrets and configuration
Secrets left in the code are detected, and public addresses are checked from the outside after each new version in production, then every night.
History
It is append-only, chained, and signed by a key that never leaves its vault. It can be exported and re-verified with a small script. With the Enterprise plan, it also streams to your SIEM.
Backups
They go to storage outside the cluster where the tools run, dedicated to the organization and versioned.
Access to Google Cloud and Azure
A production tool reaches them with no credential to keep, through OIDC identity federation.
Cutting access
You disable a service account, revoke a credential or a token, close a person’s sessions or remove someone’s access, at any time.
In case of intrusion
The while team can suspend builds and new versions in one action, and revoke access, without shutting down the tools that are running.

Questions

Who can raise a project’s budget?

A person raises it, never an AI.

Can one tool read another tool’s data?

Only the data of another tool in your organization, in production, read only, and with the approval of a person who manages that data.

What if my AI is compromised?

You can disconnect it from while at any time, and its tokens stop working.

How do I report a vulnerability in while?

Write to security@while.sh. The address is also in the site’s security.txt file.

Send your security questions to the while team

The while team’s address

hello@while.sh

Talk to the while team

To try it, give this message to your AI:

Read while.sh/agents and help me get started with while. Copy the message