Enterprise-grade depth without the “upload your entire repo” shortcut. Every workspace ships with the full security stack active: safety is never optional.
We built Patcherly around one constraint: a breach of our systems must never compromise your code. The rest follows from there.
We do not mirror your full repository. Pre-apply backups stay on your host. The AI sees limited error context for a given analysis—not your whole codebase. Suggested patches may be stored so you can review and approve them.
Not toggles you have to remember to switch off. Things the system is simply not built to do. Your code stays yours.
No full-codebase upload. Analysis uses limited excerpts; suggested patches may be stored for review.
Connectors pair with short-lived OAuth. They watch for errors, not your whole tree.
The AI receives scrubbed error context, not your codebase.
Snapshots before a patch live on your host only. We never warehouse them.
We do not hold production credentials to your servers.
Inference only. No training or fine-tuning on your data.
All of them switch on the moment you connect your first site, on every plan. There is no security tier to upgrade to.
Short-lived OAuth for connectors, rotating refresh, and per-site signed requests with replay protection. Tokens are bound to one workspace and one site.
Errors, proposed patches and paths are checked before they travel and again before they apply. Malformed or unsafe payloads are stopped early.
Not a hosted copy of your repo. Connectors send limited error context and apply best-effort secret redaction before data leaves your environment.
Before a fix touches a file, the connector snapshots it on your host with integrity checks. Those backups stay with you.
Workspaces are isolated from each other. APIs and login are rate-limited so abuse in one place cannot spill into another.
TLS on browser and connector traffic, plus transport and browser security headers on the dashboard.
Hardened browser headers, short-lived sessions, and optional TOTP two-factor on your account. No client-side secret storage.
Suspicious probing triggers blocks and audit logging. Prompt-injection hits can quarantine a draft fix; repeated hits can pause the site until you release it.
Error logs are attacker-controlled text. Prompt-injection checks run by default: suspicious context and drafted fixes can be quarantined so they cannot be approved or applied, and repeated hits can pause the site until you release it.
What we hold, what we do not, and where the legal text backs it up.
Third-party AI is used for inference only: draft a fix from scrubbed error context. Your code is never used to train foundation models (see Terms §2.5 and Privacy — Analysis and AI). Providers are listed under subprocessors.
| What | Where / How | Legal reference |
|---|---|---|
| Account data & metadata | Account/profile data and operational error/fix data are stored separately; operational records use pseudonymous identifiers. Core databases run in the EU (Frankfurt); some subprocessors process limited data outside the EEA under SCCs and similar safeguards. | Privacy §7 Privacy §5 Subprocessors |
| Source code & full repository | We do not mirror your repository. Limited excerpts may be sent only for error analysis when the connector transmits them as Service Data. | Privacy — what we do not collect DPA §3 |
| Error payloads & log context | Official connectors apply best-effort secret redaction before data leaves your environment. It cannot catch every format—never log secrets. | Privacy — Secrets and logging Privacy — Analysis and AI |
| Pre-apply file backups (connector) | Taken on your host before a patch applies, with sha256 integrity checks. Never warehoused by Patcherly. Retention is yours to manage. | Subprocessors — on-site backups |
| Patcherly application database backups | Our own disaster-recovery backups of Patcherly service databases only—not your source tree or on-server pre-apply snapshots. Encrypted in transit and at rest where applicable; pruned on a schedule. | Privacy §7 Privacy §6 |
| AI inference | Third-party AI providers receive only the error context needed to draft a fix. Service Data is not used to train their foundation models. | Terms §2.5 Privacy — Analysis and AI |
| Data Processing Agreement | Contractual processor terms (GDPR Art. 28) accepted with the Terms of Service. | DPA §1 (roles) DPA §2 (instructions) |
| Subprocessors | Published list of providers, purposes, data categories and regional notes. | Subprocessors (table) DPA §5 |
| Account deletion & export | Self-serve export and delete in the dashboard. Hard deletion from primary systems within up to seven business days (longer if law requires). Our DR backups may retain residual rows until they age out—not your connector file backups. | Privacy §6 DPA §9 |
Already using Patcherly? Security guides in the Help Center
Email [email protected] with [SECURITY] in the subject for vulnerability reports (coordinated disclosure; we aim to reply within 24 hours). Our security.txt is at /.well-known/security.txt.
Need the full technical security overview for a vendor review? Ask the same address — we send it upon request.
Last updated: September 2026. Material changes are announced in our changelog and to the responsible disclosure list.
If this model fits how your team works, the fastest way to check is to point it at one real site. Every guarantee on this page applies from the first minute, free plan included.
Full Pro plan for 30 days. No credit card. Free Personal plan after that, for as long as you like.