LangWatch provides evaluations for language models and tests for AI agent applications.
rehydra
rehydra scans LLM prompts for accidental personal information before they reach a model.

About rehydra
rehydra is an open-source SDK that scans LLM prompts for personal information before they reach a model. It detects structured data such as emails, phones, IBANs and credit cards with regex, and names, organizations and locations with an on-device NER model, then replaces values with stable placeholders that persist across a session and restores originals from an encrypted map for file writes, commands or final answers.
It can be used as a local proxy for AI coding tools, a Chrome extension for ChatGPT and Gemini, an OpenCode plugin, and an embeddable library for Node.js, Bun and browsers. The proxy requires an Anthropic API key and does not support Claude Max/Pro OAuth subscriptions through proxies, and the OpenCode plugin does not scrub the separate session-title request.
Key features
- Detect structured PII with regex patterns (emails, phones, IBANs, etc.)
- Detect names, organizations, locations with on-device ONNX NER model
- Replace PII values with stable session-persistent placeholders
- Restore original values from encrypted map for tool actions
- Local proxy for OpenAI/Anthropic-compatible clients including Claude Code, Cursor
- Chrome extension to anonymize ChatGPT and Gemini prompts in browser
- NER detection runs on-device with no cloud calls
- Stable placeholders keep entity relationships across messages in a session
- Original values are stored encrypted with AES-256-GCM for restoration
- Proxy requires an Anthropic API key and does not support Claude Max/Pro OAuth subscriptions
- OpenCode session-title request bypasses plugin hooks and can send first message without scrubbing
rehydra at a glance
- Type
- Open-source project
- Pricing
- Not confirmed. Check the site
- Open source
- Yes · Source code
- Language
- English
- Last checked
- Oct 5, 2026 · Online when checked
- Listed
- Oct 5, 2026
rehydra alternatives
Tools and sites you could use instead, ranked by how closely they match.
EvalView regression tests AI agents by snapshotting behavior and comparing tool calls in CI.
Greywall provides a deny-by-default sandbox for AI coding agents, with kernel-enforced isolation on Linux and macOS.
Prediction Guard provides a self-hosted platform for applying NIST and OWASP policies to models and agents.
Cleanlab Detect scores AI responses for trust and accuracy and flags issues such as hallucinations or missing context.
Promptfoo helps developers build and test AI applications by identifying security vulnerabilities during development.
Popular in AI Code Review & Testing
Well-known listings from the same topic.
rehydra FAQ
How does Rehydra handle PII differently from permanent redaction?
Does the proxy work with Claude Max/Pro subscriptions?
Does the OpenCode plugin scrub all requests?
Tell us and an editor will check it.
Claim the listing and suggest updates yourself. An editor checks every claim first.












