AgentRail Founding Testers

AgentRail is collecting a small evidence set from individual developers before hosted onboarding. The goal is to learn whether the SDK, MCP reader, and local Context Relay help real AI workflows preserve useful context and forensic evidence.

Who should test

  • You use Codex, Claude, or another MCP-capable local AI workflow.
  • You can test one repository or agent workflow without sharing private source, prompts, credentials, or payload data.
  • You are willing to report what was useful, confusing, slow, or not worth keeping.

What we measure

The tester loop records practical checkpoints instead of traction claims. These phrases are intentionally plain so every interview can be compared consistently.

  • installation completed
  • first Context Pack created
  • returned within seven days
  • uninstall reason

Privacy boundary

Do not share source code, private prompts, credentials, API keys, or proprietary payloads. Feedback should describe workflow category, install outcome, context quality, latency perception, and whether you would keep or uninstall AgentRail.

Apply to test

Public intake is controlled by NEXT_PUBLIC_AGENTRAIL_TESTER_INTAKE_URL. If it is not configured, AgentRail shows an honest unavailable state instead of a dead button or fake form.

Tester intake is being prepared. Use the About page for current project status until the public issue form is configured.

Before applying

Review the guided sample first. It explains the Trace Rail, Evidence Drawer, and Action Ledger using synthetic data only.

Open guided sample