For developers working with AI who protect their workstation.
- API keys
- Bearer & JWT tokens
- AWS, GitHub keys
- Private keys (PEM)
- Crypto wallets
- IBAN
Protect your secrets from AI agents
AI sees too much. LocalGuard hides what shouldn't end up in logs, context windows or vendor metrics. Works with any AI agent — from Claude and ChatGPT to your own API integrations.
API keys, Bearer & JWT tokens, AWS keys, GitHub tokens, private keys (PEM), crypto wallets, IBAN — all protected in the free tier.
This dashboard screenshot comes from a live proxy run with real secrets intercepted in real time.
Every day, developers and teams accidentally send sensitive data to LLM providers. The consequences are real and immediate.
One pasted error log can expose your OpenAI, AWS, or Stripe keys. Once sent, you lose control over who can access that data.
Credit card numbers, IBAN codes, and crypto wallet addresses in chats, logs, and AI requests are transmitted to and stored by LLM providers — often with retention policies you don't control.
Names, emails, phone numbers, and addresses from databases and logs that you paste into AI chats or analysis requests. GDPR violations, trust loss.
Database passwords, SSH keys, JWT tokens, Bearer tokens in code snippets you ask AI to debug.
LocalGuard sits between your tools and LLM providers. It masks sensitive data before sending and restores it when the response comes back. You work as usual — but your secrets never leave your machine.
"Fix this code: conn = psycopg2.connect(host='db.acme.io', password='Sup3rS3cret!', user='admin') and send email to john.doe@acme-corp.com using key sk-proj-abc123xyz789"
"Fix this code: conn = psycopg2.connect(host='db.acme.io', password='<REDACTED:password_in_text:1>', user='admin') and send email to <REDACTED:email:1> using key <REDACTED:api_key:1>"
Here is what LocalGuard catches in a real work session with Codex and Claude Code.
The metrics above and the screenshots below come from the same proxy run. Tap any image to inspect the full-size evidence.
Two-stage detection: fast regex patterns + optional ONNX AI model for names, locations, and context-aware entities.
Items marked "Free" work without a license. "Pro" items require a paid license.
LocalGuard was built by people who worked on large banking systems and business-critical software. They understand what confidential data means in everyday life, what happens when it leaks, and why protection has to work before a request ever leaves your machine.
Your data never leaves your machine for scanning. No cloud, no analytics, no third-party servers. Everything runs on your hardware under your control.
LocalGuard has already been used to protect business staff who work with confidential operational, client, and financial information every day.
The Rust core keeps scanning fast, predictable and light on resources. Sensitive data is processed and stored on the same machine where the work happens — nowhere else.
For developers working with AI who protect their workstation.
≈ $3.25/month · billed once a year
For those processing real client data.
Everything in Free, plus:
For teams with server deployments and data policies.
Everything in Pro, plus:
Yes. The proxy runs locally on your machine, intercepts HTTP requests and strips sensitive data before they reach LLM providers.
No. Regex scanning takes under 50ms per request, including the optional AI model. You won't notice any difference.
Yes. Any tool that sends HTTP requests to an LLM API can be protected. Just change the API endpoint to point through the local proxy.
Absolutely. Placeholders like <REDACTED:email:1> preserve the semantic structure. The LLM knows there was an email there, it just doesn't see the actual address.
No. Everything runs locally on your machine. No cloud services, no telemetry, no data leaves your computer for analysis.
The source code is closed, but for enterprise customers we work with security experts and provide everything they need for an independent review.
The screenshots on this page come from a real local session where secrets appeared in more than 70% of requests. Even careful users paste error logs, stack traces, and config fragments under pressure. Automation catches what attention misses.
For most individual developers — yes. The free tier detects and redacts API keys, tokens, private keys, crypto wallets, and IBAN codes. If you also need password, credit card, email, phone, or AI-powered name/location detection, upgrade to a paid license.
If your machine fingerprint hasn't changed, open LocalGuard, copy the fingerprint from the License panel, then go to Recover and paste it there. You'll get a freshly signed license key for that machine — paste it back into LocalGuard's Activate License field. The original expiry date is preserved. If your fingerprint did change (e.g. hardware swap or major OS update), you'll need to buy a new license.
Free and Pro are designed for a single user on a single machine. Enterprise lets you run the proxy on your organization's servers and grant simultaneous access to many employees without installing the app on every workstation — so you can enforce security policy on LLM-bound data company-wide.
Start free — API keys, tokens, and crypto detection included. Download in 2 minutes, no license needed. Upgrade anytime for passwords, credit cards, PII, and AI model.
One pasted log, client record, or configuration fragment can expose money, access, and confidential work. Then you no longer control who sees the data or where it is stored.
This is not theory. Shared AI chats have already appeared in Google. In our local session, sensitive data showed up in 70%+ of outbound AI requests.
Emails, phone numbers, passwords, and private details can leave with one careless paste.
Cards, banking data, crypto wallets, and payment keys can turn into direct financial loss.
Client data, contracts, internal notes, and access tokens can stop work, damage trust, and cost someone a job.
Tell us about your team and we will prepare an email to support@localguard.me.
This opens your email app and prepares a message to support@localguard.me.