Free forever · Open source

Keep secrets out of AI.

Catch API keys, tokens and credentials before they reach an AI assistant, in your terminal and your browser. On-device, zero network.

$ npm i -g @sbr0nch/contextia
Read the docs View on GitHub
contextia · live demo on-device
Connection stringpost…5432AWS access keyAKIA…MPLEGitHub tokenghp_…aaaa
3 secrets found

↑ Live secrets found. Hit Redact all to strip them. It all runs in your browser.

One engine. Four surfaces.

The same on-device secret detection, wherever you work with AI.

Terminal & proxy

Scan a diff, or proxy any AI agent (Claude Code, Cursor, Windsurf, aider) so secrets are redacted or blocked before they leave.

$ contextia run -- claude

Claude Code plugin

Blocks a prompt the moment it contains a secret, natively, before it reaches the model.

$ /plugin install contextia@contextia

Browser extension

Coming soon

On-device secret detection across ChatGPT, Claude, Gemini, Copilot, Perplexity and more, redact, allow or block.

Engine

The same 58-detector engine, embeddable in your own tools.

$ npm i @sbr0nch/contextia-engine

Proxy and extension can redact; the plugin blocks, because Claude Code hooks can't rewrite a prompt.

Zero network requests.
Everything runs on your device.

MIT, open source · 58 detectors · no accounts, no telemetry.

GitHub· npm

Get started

Terminal

CLI + local AI-DLP proxy. Live on npm.

$ npm i -g @sbr0nch/contextia
View on npm →

Claude Code plugin

Native block, no proxy. Self-contained, no CLI needed. Add the marketplace, then install.

$ /plugin marketplace add sbr0nch/contextia
$ /plugin install contextia@contextia
Plugin docs →

Browser

Submitted to the stores. Approval pending.

Chrome Web StoreComing soon
Firefox Add-onsComing soon
Build from source
For teams · Enterprise

Run Contextia across your whole org

SentriKat is the security platform where teams manage vulnerabilities, compliance and now AI-DLP. It's the managed home for Contextia — run it from a console instead of device by device.

  • Central rollout — push to every browser and machine from one console
  • One policy — set allow / redact / block rules once, org-wide
  • Audit & compliance — a searchable log of every secret caught
Explore SentriKat Same team behind Contextia.
SentriKat consoleorg-wide

Protected endpoints

ChatGPT Claude Cursor Terminal

Audit trail

  • AWS access keyredactedCursor
  • GitHub tokenblockedChatGPT
  • DB connection stringredactedTerminal
Scanned with SentriKat
Continuous vulnerability scanning
SBOM