Solo founder building @inariwatch. AI fixes your production incidents — fleet of agents, cryptographically verified, shipped end-to-end. Rust · TS · Go.
Spent 6 months on this: an AI agent for production engineering where every action is cryptographically signed.
Apply fix → diff lands → click the receipt chip → see the chain.
The first AI you can put on-call without losing sleep.
Inari Live. Soon. 🦊
Your on-call engineer gets paged at 3am.
They open the alert. Stack trace. No context.
30 minutes debugging what even caused it.
@inariwatch/capture sends the AI everything —
git commit, request, breadcrumbs, user — the moment it crashes.
Fix in minutes, not hours.
We're shipping this exact pattern for production remediation at InariWatch.
Currently at 20% canary, full rollout in a few days once the metrics close: hypothesis diversity ≥ 3 per session + success rate ≥ single-agent baseline.
Built clawsweeper, which runs 50 codex in parallel around the clock, scans issues/prs deep and closes what is already implemented or what makes no sense.
Closed around 4000 issues today, a few thousand are in the pipeline. (rate limits are rough) https://t.co/AiNNDcvGke
🚀 I’m building InariWatch
The first monitoring tool with a full closed-loop repair cycle
detect → diagnose → replay → fix → CI → deploy → monitor → revert
Not just alerts.
Not just dashboards.
Your app actually heals itself.
⚙️ How it works
• Connects to Sentry, Vercel, GitHub, Datadog
(or use our SDK — 1 line)
• AI diagnoses root cause on every alert — for free
• Session replay shows the full causal chain in ONE timeline
click → API → DB query → error
(frontend + backend aligned, no tab-hopping)
• Click “Fix It” →
clone repo → generate fix → run tests → open PR
• 17 safety gates before anything merges autonomously
• Post-merge monitoring auto-reverts if regression is detected
💡 All AI features are free
🔑 No API key needed
👨💻 Solo dev, building in public
🔥 Beta is open
👉 https://t.co/50RFxYZTHt
POV: tu app rompe en producción a las 3AM. 😴💥
Antes de que despierte tu on-call, InariWatch ya:
1️⃣ Diagnosticó la causa raíz
2️⃣ Generó el fix
3️⃣ Pasó 17 gates de seguridad
4️⃣ Firmó criptográficamente el receipt
5️⃣ Te muestra VISUALMENTE cómo queda la app con el fix aplicado
Y tú… sigues durmiendo. 🛌
Demo pronto 🔥
@edgaru_angel@BBVA_Mex Jajaja gracias 🙏 ya lo estoy imaginando. Enterprise con fixes autónomos + preview visual + prueba criptográfica = bancos pueden dormir tranquilos. Se viene pronto 👀
I'm building InariWatch — the first monitoring tool with a full closed-loop repair cycle.
detect → diagnose → fix → CI → deploy → monitor → revert
Not just alerts. Not just dashboards. Your app actually heals itself.
How it works:
- Connects to Sentry, Vercel, GitHub, Datadog (or use our SDK — 1 line)
- AI diagnoses root cause on every alert, for free
- Click "Fix It" → clones repo, generates fix, runs tests, opens PR
- 11 safety gates before anything merges autonomously
- Post-merge monitoring reverts if regression detected
All AI features are free. No API key needed. Solo dev, building in public.
Beta is open: https://t.co/8yCF3Er3Nm
Session Replay is live on InariWatch 🎬
• DOM + console + network + Web Vitals — all in one timeline
• Backend I/O synced to the same scrubber
• Rage + dead-click detection
• AI-narrated chapters
• One click → PR with the fix
https://t.co/X3rixBrD8W
Building my own Managed Agents for InariWatch
The AI spins up a Docker container, clones your repo, explores the code, writes the fix, and runs tsc + build inside the container — only pushes code that compiles.
No broken pushes. No CI retries.
The AI verifies its own code before pushing.
npx @inariwatch/mcp init
Your app breaks at 3am. InariWatch catches it, reads the session replay, writes the fix, runs 11 safety gates,
deploys to staging, and ships — or drafts a PR if anything fails.
One command connects all your AI tools.
https://t.co/hEQ0GYZ7tX
Your AI coding tool is guessing what broke in production.
InariWatch gives it the full picture — stack traces, git context, I/O recordings, breadcrumbs — before you even ask.
One command. Any AI tool.
npx @inariwatch/mcp init
https://t.co/hEQ0GYZ7tX
Your AI coding tool now knows what's broken in production before you open your editor.
17 MCP tools. One URL. Any AI tool.
Claude Code, Cursor, Windsurf, VS Code, Codex, Gemini — all connected.
npx @inariwatch/mcp init
https://t.co/hEQ0GYZFjv
Your AI coding tool now knows what's broken in production before you open your editor.
stack trace. logs. root cause. all there.
one URL. any AI tool.
https://t.co/hEQ0GYZFjv
Just shipped @inariwatch/capture v0.2.0
One line of config:
init({ dsn: "...", substrate: true })
Your app now records every HTTP call, DB query, and file operation. When an error happens, the AI doesn't guess — it sees exactly what your code did.
Error capture + full I/O recording. One import.
npm install @inariwatch/capture
https://t.co/EHykWVGp7y
Every monitoring tool stops at "here's your alert."
InariWatch reads your code, writes the fix, pushes a branch, waits for CI, and opens a PR.
From error to merged fix. 2 minutes. No human needed.
Open source. https://t.co/hEQ0GYZFjv
@1Umairshaikh Building InariWatch — AI that fixes your production errors and opens a PR while you sleep.
When something breaks: reads your code → writes the fix → waits for CI → opens a PR.
6 safety gates. Open source. AI included free.
https://t.co/hEQ0GYZFjv
@TheZilisch Same here. Building in public is the only channel that works at zero scale.
Ship → share → listen → repeat. Everything else is noise until you have the product people want.
Good luck with PingSentry — rooting for you. 🦊
Tools InariWatch replaces:
Sentry → @inariwatch/capture ($0)
PagerDuty → on-call rotations, escalations ($0)
Pingdom → uptime monitoring ($0)
Statuspage → /status page ($0)
BYOK — bring your own AI key. We never proxy it.
100% open source. MIT license.