One governed home for AI prototypes. Your identity provider, your storage, your servers.
Your designers build with different AI tools. They publish into one workspace, with no repository or deploy pipeline needed to get a prototype reviewed, behind single sign-on and directory sync, an audit log you can export, your own S3-compatible storage, or a self-hosted edition that runs air-gapped.
Every designer builds with a different AI tool. Your prototypes need one governed home.
Each designer builds with a different AI tool, and the results end up nowhere in particular.
Publish from Claude Code, Codex, Cursor or Lovable: a static build as a folder, a zip or a single HTML file, or let the coding agent publish it. Every prototype lands in one workspace, in folders per team or product, with a thumbnail and its status, and ⌘K finds any prototype or comment.
To share a prototype, a designer has to go through a repository, a build pipeline and a preview environment.
Drop the build into the workspace, run crit publish, or let the agent do it. Nobody needs a repository or a pipeline to get a prototype reviewed, so product managers and stakeholders without repository access can comment. Where engineering wants GitHub or GitLab, a prototype carries its repository link and each version its commit, branch and pull request, and you choose who can see the code.
Sign-in has to go through our identity provider, and we need to know who did what.
Connect SAML single sign-on: people from your verified email domain go straight to your identity provider, and you can require it. SCIM keeps people in step with your directory, prototypes are open to workspace members by default, guests see only the folders you share, and the audit log records every change.
Our data has to stay where we put it.
Bring your own S3-compatible bucket, with EU region pinning. Or run critlab on your own servers, air-gapped if you need it.
The same review loop. Inside your controls.
Moves to the next tab every few seconds. Choose a tab to stop.
A terminal: crit publish puts the prototype online and prints its share link. Tina and Leo comment on two elements. Then, in Claude Code, the agent pulls both comments over MCP, commits the fix, publishes version 2 to the same link and resolves both comments.
- 1
Sign in through your identity provider
SAML single sign-on by verified email domain, with SCIM keeping people in step with your directory.
- 2
Publish from your agent
With a key that carries only the permissions you grant. Each version runs on its own isolated origin.
- 3
Review with the right people
Members, and guests who see only the folders you share. Each prototype is open to workspace members only, a password or a link.
- 4
Keep the record
Every change is in the audit log, ready to filter and export.
Built for your security review. And for the teams who use it.
Single sign-on.
SAML with your identity provider. People from your verified email domain go straight to it.
Directory sync.
SCIM provisioning: people join and leave with your directory.
Roles and guest access.
Owner, admin, member and guest. Guests see only the folders you share with them.
Audit log.
Every change on record. Filter by person, folder or prototype; export CSV or JSON.
Custom share domains.
Share links on your own domain.
Your own storage.
Bring an S3-compatible bucket, with EU region pinning.
Sandboxed prototypes.
Every version runs on its own isolated origin, and nothing is ever indexed.
Scoped keys for agents.
Each agent key carries only the permissions you grant it.
Self-hosted.
The same codebase on your own servers. Docker Compose or Helm, air-gapped if you need it.
Your staging site, too.
One script tag brings the same comments to a site you already host.
No account for reviewers.
Share an open link or add a password, and reviewers need no account. Or keep it to workspace members.
Inspect and code.
Inspect an element's styles and spacing, read the files and see what changed between versions. You decide who sees the code.
Questions
Anything else? Write to us at support@critlab.app