About Patcherly

Where AI tools & IDEs end, Patcherly begins.

Copilot, Codex and Claude Code author code. Sentry, Datadog and Bugsnag see what broke. Neither closes the production loop — triage, draft, approve, verify, roll back. That gap is where Patcherly lives.

We don’t replace your IDE assistant or your observability stack. We sit between them, on your infrastructure, and finish the work they start.

Read the research

Our mission

Make production debugging boring — in the best way. The kind of fix that resolves itself before the on-call engineer has finished their coffee, with a full audit trail and a one-click undo.

Approval-first by default
Code stays on your server
Roll back in one click
EU-hosted, GDPR-aligned

The cost of debugging today

Debugging is the silent tax on software.

Three numbers nobody puts on a roadmap: the time it eats, the bill a small team pays, and the dent it leaves in the global economy.

~41%

of developer time is spent on bugs and maintenance.

Not shipping new features and improving workflows, but drowning in debugging, refactoring and tech debt.
Stripe Developer Coefficient Report →

~€8,500 /year

spent on manual debugging at ~90k visits/month.

Modelled developer hours lost to triage and patching, before Patcherly enters the picture.
Patcherly Research Findings →

$2.41 Trillion

lost by the U.S. economy each year to poor software quality.

The annual cost for businesses, due to operational failures, technical debt and failed projects combined.
CISQ Report →

The research behind Patcherly

What changes when AI sees the whole stack.

Our hypothesis — that AI given full system and application context can resolve a meaningful share of production bugs autonomously, under human approval — was tested across one million simulated sessions in the dissertation that became this product.

~3.4×

More errors auto-resolved.
~18% resolved with enriched context vs. ~5% without — same error pool, same review process.

~57%

Less time on debugging.
Resolution time fell 57% in the base scenario. At 100k sessions/day, the modelled saving exceeds €250,000/year.

~12%

Smaller manual burden.
Errors needing a human dropped from ~94% to ~82% — roughly one in eight reclaimed for full automation.

Methodology, full charts and tables: Research findings →
Martelli, J. (2025). Leveraging AI for Automated Self-Healing in Web Applications. OPIT BSc Modern Computer Science thesis.

Principles

How we make decisions.

Four constraints shape every Patcherly decision. Not features. Not toggles. Non-negotiable.

Approval before action

A patch that breaks something is worse than no patch. We’d rather pause and confirm than be fast and wrong — which is why approval-first is the default, not an option.

Privacy by architecture, not policy

We can’t leak what we don’t have. The connector ships only what analysis needs; sensitive values are stripped before they travel; backups never leave your machine. Not settings — how the system is built.

Reversibility by design

Every automated action has an undo. Rollback isn’t a fallback we bolted on — it’s the constraint we designed around. If a tool can’t undo its mistakes safely, it has no business in production.

Transparency as accountability

Who approved, when it ran, what the confidence was, how it ended — every step leaves a record. That record belongs to you, so you can audit it, explain it to a client, or learn from it.

Who’s behind it

A solo founder with skin in the game.

Patcherly is built by Jany Martelli, a computer scientist who has shipped and debugged production code for over two decades. Every feature is grounded in a real night spent firefighting something that shouldn’t have broken.

The founder is also the primary on-call for Patcherly itself. If it isn’t safe enough for us, it doesn’t ship for you.

Patcherly on GitHub

Made in Europe

Community

Hang out with early users. The founder answers questions there; fixes and roadmap feedback get traded daily.

Join Discord →

Special thanks to

Beta testers and design partners welcome.

Patcherly is in private beta. Early seats come with a lifetime discount, a direct line to the founder, and shaping power on the roadmap. If you’re a hands-on developer, and have opinions, we want to hear them.

Talk to the founder