SnippetVeil

Anonymize a Java or Kotlin snippet, a stack trace, or a database execution plan, before you paste it into an AI chat.

Every name your project owns — classes, methods, fields, packages, string literals — is replaced with a placeholder before the snippet leaves the editor. The structure, the control flow, and the libraries you call go as-is.

How it works

• Select Java or Kotlin code, then right-click → SnippetVeil → Copy Anonymized. The anonymized text is on your clipboard.
• Commented-out code is kept, with its names replaced. In a .java or .kt file, a comment whose text is code, such as // this.customer.setOrder(order);, comes out as // this.field2.method3(param4);: every name in it takes the placeholder that name gets in the code around it. A comment that is prose, such as a TODO or an explanation, is stripped.
• Anonymize with Preview… shows the code that will be copied beside its mapping table, with counts: 14 renamed · 3 unknown · 22 preserved · 2 comments stripped. It is the only place you can anonymize less — keep comments for this snippet, or preserve a name and send it exactly as written. Names SnippetVeil could not resolve can be preserved directly; every other name becomes preservable once you unlock the column, which warns you first and is locked again the next time you open the preview.
• Renaming a placeholder in that same table gives the model the word your question is about: double-click a Placeholder cell and Type1 becomes FilterType1. The stem is yours, the number always stays, and your real name never leaves the machine. Only names this snippet introduced can be renamed — one an earlier snippet already used keeps the name that snippet's replies are written in.
• De-anonymize Clipboard turns the placeholders in a reply back into your real names.
• De-anonymize Clipboard and Paste does the same and inserts the result at the caret, replacing the selection. It will not paste a reply it could not restore in full, so placeholders never reach your source; the clipboard keeps the anonymized reply either way.
• The mapping is per-project and stays on your machine. Placeholders are stable across invocations and IDE restarts, so a second snippet never contradicts the first.
• SQL in a Java file is read name by name where the IDE recognises it as SQL or JPQL. "SELECT state FROM billing.customers" comes out as "SELECT col1 FROM schema2.table3", and a JPQL query's entity and field names take the placeholders the class and its fields get everywhere else. A query SnippetVeil cannot read in full is replaced whole, as one "str1": SQL holding a parameter, a string or a comment, JPQL with a name that does not resolve, and every query in a Kotlin file. Which queries the IDE recognises depends on the plugins it has.
• Anonymize Execution Plan… reads an execution plan from the clipboard — PostgreSQL, MySQL, SQL Server or Oracle, in the output forms each engine prints unambiguously — and gives back the same plan with its table, column, index and alias names replaced and the values in its conditions redacted. Every cost, row estimate and timing stays exactly as printed. Copy the plan, then right-click → SnippetVeil → Anonymize Execution Plan…; it always opens the preview first. A plan SnippetVeil cannot read safely is refused, your clipboard is left as it was, and the message names the output option that works where there is one. A plan's names are not added to the project's mapping, so the same plan pasted twice comes back under different numbers.
• Anonymize Stack Trace… reads a Java or Kotlin exception trace from the clipboard and gives back the same trace with the package, class, method and file names your project owns replaced by the placeholders your snippets already use, so its frames line up with the code you sent beside it. Every line number stays as printed, and JDK and library frames keep their names unless the library is your organization's; the module or classloader prefix on a frame, such as java.base/, is removed. Every exception message and the thread name become str placeholders, and a name the IDE cannot find becomes Unknown. Copy the trace, then right-click → SnippetVeil → Anonymize Stack Trace…; it always opens the preview first. A clipboard holding anything besides the trace is refused, and your clipboard is left as it was.

No network

SnippetVeil makes no network calls. No networking code — enforced on every pull request, scanned in every release build, and public from the first commit.

What SnippetVeil does not hide

• Your stack. Frameworks, libraries, the identity provider, the persistence layer — preserved by design, because they are what make the snippet answerable.
• Your structure and your algorithm. Control flow and logic survive intact. If your proprietary value is the algorithm, SnippetVeil does nothing for you.
• The prompt you type around the snippet. SnippetVeil anonymizes the code, not the sentence above it.

What it does not preserve

• The text inside your strings — a string literal in your own code becomes "str1" → retype the value that matters into your prompt.
• Your comment prose, by default.
• Your exception messages → retype the message that matters into your prompt.

Non-goals

Not a secret scanner. Anonymizing code is unavailable outside supported source files; an execution plan or a stack trace is read from the clipboard, wherever you are. The reversal actions are not gated by language.

Source

Apache-2.0. Source and issue tracker: https://github.com/snippetveil/snippetveil · Threat model, including the limits of the no-network checks: https://github.com/snippetveil/snippetveil/blob/main/THREAT-MODEL.md

---

About this repository

An IntelliJ IDEA plugin, for IDEA Community, IDEA Ultimate and Android Studio.

Status: published. SnippetVeil on the JetBrains Marketplace — Settings → Plugins → Marketplace, and search for SnippetVeil. Every release is signed, and the distribution zip is attached to the matching GitHub Release for anyone who would rather install it from disk and check the bytes first.

This repository is public from its first commit, deliberately — a continuous commit history is the evidence behind the no-network claim, and it can only be accumulated by starting early.

Work in progress is tracked in this repository's issues.

Every fixture in this repository is synthetic. The anonymiser is exercised against real codebases by a local instrument that is never run in CI and writes its findings outside the repository tree — and a bug it finds earns a synthetic fixture reproducing the shape, never the real code that revealed it. Running the anonymiser over a real codebase and committing the output would be exactly circular: if it leaked, the leak would be in a public repository, permanently. See CONTRIBUTING.md.

The demo/ directory is the sample project the Marketplace screenshots are shot from. It is not part of the build and is not in the distribution — see demo/README.md.

Documents

• THREAT-MODEL.md — what SnippetVeil conceals, what it discloses, and what defeats it.
• PRIVACY.md — what SnippetVeil stores and where. A factual statement, not a policy.
• CONTRIBUTING.md — building, the trust checks, the DCO, and the inbound dependency policy.
• LICENSE and NOTICE — Apache-2.0.

Support

SnippetVeil is free and Apache-2.0, and anonymizing your code is not something you will be asked to pay for.

If it saves you time and you want to fund the maintenance, there is a Sponsor button at the top of this repository, and the same link on the Marketplace listing. It is optional and it unlocks nothing — no features, no priority, no private builds. The plugin itself never asks: it makes no network calls, so it has no idea whether you sponsored it.

Two things help more than money. A bug report carrying the shape of the code that broke — synthetic, never your real source (CONTRIBUTING.md). And a rating on the Marketplace, which is one of the few signals this project gets, because it collects no telemetry.