Built for BrowserPod variants - LLM-assisted proof

Prompt once.
Run verified variants.

ForkLab lets AI generate code, then runs it inside BrowserPod before you trust it.

Enterprise TrustGenerating
> provider=gemini
> wrote public/index.html
> wrote public/styles.css
Startup Conversion Installing
> boot BrowserPod
> npm install
> waiting for Portal callback
✉ Developer Minimal✓ Portal ready
> FORKLAB_VARIANT_NONCE matched
> Express static server listening
> BrowserPod Portal captured

Final Demo Modes

Mode 1

Arena

Select sandbox issues from a simulated GitHub repo and run parallel BrowserPod branches.

Primary Demo

Mode 2

Arena Live

Prompt for frontend variants and run them directly in BrowserPod with immediate terminal and visual feedback.

Mode 3

Workbench

Enterprise agent workflow with proof. Shows the steps of prompting, generating, running, and verifying.

Zero Trust Execution

Software Security First

Untrusted AI-generated code should never run directly on your host machine. ForkLab treats all LLM outputs as potentially malicious or broken.

Isolated ephemeral containers per branch
No access to local filesystem or secrets
Automated rollback on failure or timeout
Security.log
[SYS] Initiating Pod Isolation Sequence...
[SYS] Mounting virtual fs: /tmp/pod_env_a4f2
[NET] Restricting outbound traffic to whitelist
> Writing generated variant files...
[SEC] Host filesystem unavailable to sandbox code
> Nonce proof returned from BrowserPod.
[SYS] Shredding container instance a4f2...
[SYS] Host machine remains secure.