NanoNeuron
AirlockPaid API · live

Your agent reads things you didn't write.

A support ticket, a web page, a PDF, an email. Any of them can carry a sentence aimed at the model rather than at you. Airlock sits between your agent and that content — scoring what comes in for prompt injection, and checking what goes out for credentials and personal data.

Recorded output from the detector, replayed. The live scanner runs on whatever you paste into it from your own credits — nothing stored, no free tier.

PostMortem AI · for DevOps, SRE, and freelance consultants

Postmortems that cite their evidence.

Skip the Friday-afternoon postmortem slog — get an AI-drafted writeup where every claim traces back to a real, recorded entry. Anything the evidence doesn't support is marked unsupported, never invented.

What “grounded” actually means

Kept — cites real evidence

“Latency rose from 180ms to 4200ms starting 14:04 UTC, correlated with deploy #482.[2]

Dropped — no citation exists

“Not established by the recorded evidence.”

(what a claim like “this cost the company $40,000” becomes, if no evidence entry says so)

How it works

1

Record evidence

Alerts, deploys, metrics, human notes, customer reports — timestamped entries, not a blank text box.

2

Generate a grounded draft

The model cites which evidence entry backs every claim. Code checks every citation independently before anything is stored — a claim with no valid citation is replaced with a fixed unsupported marker, never kept.

3

A human approves, then it publishes

Publishing always records a real, named approver. The database itself refuses to mark a postmortem published without one.

Integrates with

LinearLinear
Google GeminiGoogle Gemini

Welcome back

Log in to your account.

No account yet?

What this isn't
  • Doesn't auto-publish anything — publishing is always a deliberate, named human action.
  • Doesn't estimate cost, revenue, or customer-impact figures the evidence didn't state.
  • Has real, documented setup for PagerDuty and Datadog webhooks — not for other monitoring vendors yet, though any tool that can POST JSON can still use the generic webhook.
  • Doesn't support teams or organizations yet — each account is a single user's own incidents.