10 repository checks over 14 days, no card. You sign in with Google or Microsoft — we receive your email address and nothing else.
Sign in to redeem your voucher.
Or choose a store: Chrome Web Store · Microsoft Edge Add-ons
Chrome shows a "developer mode extensions" notice while it is loaded this way; that goes away when you install from the Chrome Web Store instead.
Repositories you chose to permanently allow from a blocked page. They apply on every browser you sign in to. A repository later confirmed malicious is still blocked — an allow is a convenience, never an override of that.
A deep scan reads the files of a repository, a gitlab.com project, an editor extension package (vsx/publisher/extension) or an agent skill in an isolated sandbox with no route to the public internet. You can start one here, or from a blocked page when a download is stopped.
Credentials your deep scans found committed in a repository, by repository and file. RepoGates never stores the secret itself — only where a scan found one. Rotate the key at its provider first: deleting it from the file does not take it out of the history, and it works until it is revoked. Then mark it here.
If you publish an agent skill or a plugin on GitHub, RepoGates can give it a public page and a README badge showing its current verdict. What is published is the verdict, the checks and the dates — never your name or e-mail address. You can remove it at any time.
If you maintain a repository on GitHub, choose A repository: the same page and badge, for the repository.
Commit this to your skill's README on the default branch, then press Verify. The code in it works once.
The runs of /repogates:checkup and the repogates-checkup command line on your machines, newest first — the score, grade and action, each agent as inventoried or detected only, and the findings with their fixes. A run sends RepoGates identifiers, never contents; what is kept here is the same. The newest thirty are kept, and all of them go with the account. How to read a check-up.
If you already pay for a scanning service, RepoGates can use it with your own credential, under your own terms. Your key is encrypted, is never shown again after you save it, and is used only for your scans. Each service is listed with exactly what it is sent — read that before you enable one.
A deep scan of a repository over the size cap cannot read every file, and a scan that did not read everything never returns PASS. Connect an AI provider and it is shown the text the engines read and the paths of everything left unread. It lists the unread files it thinks need opening, and it can raise a concern or the verdict. It adds; it never takes away. It cannot clear a finding an engine made, and it cannot turn a partial scan into a PASS.
This spends your money, not ours. The key is yours: your provider bills you directly, at their rates, under your own terms with them. Roughly one call per over-cap scan. RepoGates charges nothing extra — a scan costs the same credits with the review as without it.
Read what is sent. This is the one place RepoGates sends file content anywhere: the text of the files the scan did read, plus the paths — not the contents — of the files it could not open. Nothing is executed.
Connect more than one and Anthropic is used first, then OpenAI, then Gemini. It must be an API key from the provider's developer console — a Claude Pro/Max or ChatGPT Plus subscription carries no entitlement to serve an application, and its credentials will not be accepted.
For the RepoGates Claude Code plugin and any script that asks the verdict API on your behalf. A token can read verdicts and nothing else: it cannot open this page, change your plan or delete the account. It is shown once, when you create it, and you can revoke it here at any time.
Tokens work on Personal, Premium and Team — a free account gets a named refusal, not a verdict. See plans →
Copy it now — it is not shown again. Save it from a
terminal, not through an agent:
umask 077; mkdir -p ~/.config/repogates; pbpaste > ~/.config/repogates/token
— or set REPOGATES_TOKEN in the environment.
Applications you approved on the “Connect an AI agent” page, such as Claude or ChatGPT using the RepoGates MCP server. Each can ask for verdicts and spend your allowance and scan credits. Disconnect one and it stops working within 30 seconds; it would have to ask you again. The name is chosen by the application itself — the address is what RepoGates can vouch for.
A device is a random id minted per install — not a fingerprint. Remove one to free a slot.