Tracehold Free

Detects passwords, API keys and personal data before you send them to ChatGPT, Claude or Gemini. Your prompts are analyzed locally.

Tracehold is a privacy guardian for when you use artificial intelligence. How many times have you pasted text into an AI chatbot without checking what was in it? A password, an API key, a customer email, a card number... Once sent, there is no way back. Tracehold Free analyzes what you write before you press Enter and, if it detects sensitive data, blocks the send and warns you. You get one last chance to remove that data before sharing it with an AI. Analysis happens in your browser and your content never leaves your device. Tracehold Free has no account, needs no sign-up and does not send the content of your messages to any server. Text analysis runs 100% in your browser. The detection ruleset ships inside the extension package (it is not downloaded over the network; it only changes with each new version). Besides the fully local analysis, the extension makes up to five kinds of connections to app.tracehold.com: an anonymous statistics ping at most once every 12 hours (installation UUID, date, version and browser type), a batch of anonymous scan events roughly every 2 minutes when there is activity (AI domain, action and those same anonymous fields), an optional report sent only if you press "Is this a false positive?", an anonymous service-configuration lookup at startup (URL discovery only; no data or identifiers) and a single anonymous notice on uninstall (uninstall count). None of them include the text of your messages. Your data is never used to train models. What it detects Passwords and secrets: variables like PASSWORD=, SECRET=, API_KEY=, .env files. Vendor keys: cloud provider, source-control and payment-processor API keys and similar. Tokens and certificates: JWT, private keys (-----BEGIN PRIVATE KEY-----). Connection strings with user and password (protocol://user:password@host). Credit cards (validated), IBAN, national IDs, emails and phone numbers. Where it works Tracehold Free activates on today's most popular AI chat assistants, coding assistants, writing tools and image/audio generators. The catalog grows with every version and currently covers dozens of AI provider families across general-purpose chatbots, developer and coding assistants, writing and paraphrasing tools, and creative image or audio generators. The network access list (host_permissions) is exactly the one in the manifest.json. Detection additionally recognizes a larger curated catalog of domains and only activates on those sites. Note on access: the detector is injected on every page (content_scripts with *://*/*) and unloads itself instantly except on recognized AI assistants; that is why the browser may show a broad-access warning. <all_urls> is not in host_permissions. We do not claim that the extension only accesses AI sites: we describe the design as it is. How it works 1. Install Tracehold Free and open your favorite AI assistant. 2. Write or paste your message as usual. 3. If there is sensitive data, Tracehold blocks the send and shows it to you. You decide: edit the message or send it anyway. No configuration, no buttons to press. It works from the first minute. Free version This is the free version of Tracehold, with the core detection ruleset and a limit of 3 blocks per 5-hour window. Perfect for personal use and to see how Tracehold works. After the limit, Tracehold keeps detecting and informs you, but no longer blocks. Need more? Tracehold for teams and companies The full version adds custom policies, auditing, central management, team protection and much more, with unlimited detection. Discover it at https://tracehold.com