AgentCloak is a free browser extension and desktop app that swaps real names, addresses, and account numbers with synthetic data so ChatGPT, Claude, Gemini, Grok, and Copilot never see the original values. The tool sits between the user and the chatbot, detecting sensitive fields locally and replacing them with consistent digital twins so "Peter Parker" becomes "Julio Schmidt" and "364 23rd St" becomes "401 Vanessa Curve," then restoring the real values when the reply comes back . It targets employees and individuals who already paste emails, contracts, customer records, and personal details into frontier models and want to keep using those models without exposing the underlying data.
The extension installs on Chrome, Safari, and Edge, or as a native app for Mac and Windows, and covers ChatGPT, Claude, Gemini, Grok, Copilot, and DeepSeek . Under the hood, it pairs deterministic rules with Rampart, a compact model from the U.S. government that mixes deterministic and neural detection of sensitive data, running entirely on-device and combined with AgentCloak's digital twin technology . The free release detects first and last names, email addresses, phone numbers, URLs, Social Security and tax ID numbers, bank routing and account numbers, and physical addresses across English, Spanish, French, German, Italian, Portuguese, and Dutch.
The launch matters because it takes a capability previously reserved for regulated enterprises and puts it on a consumer laptop for free, positioned by founder Peter Yared as an alternative to buying dedicated hardware for a private local model. Yared is Founder and CEO of InCountry , the data residency company behind AgentCloak, which raised $10 million in funding alongside the launch of AgentCloak for AI agent data protection . The paid AgentCloak Server tier extends the same approach to teams handling patient records and other regulated data across borders.
Comments (12)
Sign in to join the discussion.
N
Nadia Okafor25d ago
Before I take this to legal, is there a SOC2 Type II report and does the cloaking happen on-prem or in your cloud? Data residency answers determine whether procurement even opens the deck.
H
Hiro Tanabe25d ago
hot take: the real moat here isn't the cloaking, it's convincing employees to stop pasting the customer list into ChatGPT at 11pm.
S
Sasha Delgado25d ago
Distribution question: are you selling this to CISOs or sneaking in bottom-up via a browser extension? Very different motions and your landing page hedges both.
V
Vikram S.25d ago
Per-token pricing, seat-based, or flat platform fee? Asking because 'inspects every AI prompt in the org' sounds like a line item that grows faster than headcount.
M
Marisa Ionescu25d ago
Reminds me of what a couple portfolio cos tried to build in-house and shelved. If you nail the enterprise wrapper this is a very easy intro conversation.
M
mihaela25d ago
So it's a bouncer for your prompts checking IDs at the door. I respect it.
K
Koji Alvarez25d ago
Is there a proxy endpoint I can point my SDK at, or does every model need a bespoke integration? Also curious about rate limits when the cloaking layer is in the hot path.
P
Priya Raman25d ago
The tweet buries the actual demo. You had one job: show the before/after prompt side by side in the first 3 seconds.
B
Benoît Marchetti25d ago
Been circling this problem from the compliance angle at my shop. Curious how you handle re-identification on the response side without leaking the mapping table.
Z
Zara Bekova25d ago
That tweet cuts off mid-sentence at 'actual https://t.co/...' which is either the worst luck or the best engagement hack I've seen this week.
T
Tomás Verhoeven25d ago
Peter, off the record or on, are the Chinese AI services actually customers already or is that positioning? Happy to keep it background.
A
Adeola Krivov25d ago
Cloaking data before sending to third parties feels like it wants a verifiable on-chain attestation layer. Just saying, the primitive is right there.
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
AgentCloak
Cloak sensitive data at every cross-border touchpoint.
Cloak sensitive data at every cross-border touchpoint.
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
AgentCloak is a free browser extension and desktop app that swaps real names, addresses, and account numbers with synthetic data so ChatGPT, Claude, Gemini, Grok, and Copilot never see the original values. The tool sits between the user and the chatbot, detecting sensitive fields locally and replacing them with consistent digital twins so "Peter Parker" becomes "Julio Schmidt" and "364 23rd St" becomes "401 Vanessa Curve," then restoring the real values when the reply comes back . It targets employees and individuals who already paste emails, contracts, customer records, and personal details into frontier models and want to keep using those models without exposing the underlying data. The extension installs on Chrome, Safari, and Edge, or as a native app for Mac and Windows, and covers ChatGPT, Claude, Gemini, Grok, Copilot, and DeepSeek . Under the hood, it pairs deterministic rules with Rampart, a compact model from the U.S. government that mixes deterministic and neural detection of sensitive data, running entirely on-device and combined with AgentCloak's digital twin technology . The free release detects first and last names, email addresses, phone numbers, URLs, Social Security and tax ID numbers, bank routing and account numbers, and physical addresses across English, Spanish, French, German, Italian, Portuguese, and Dutch. The launch matters because it takes a capability previously reserved for regulated enterprises and puts it on a consumer laptop for free, positioned by founder Peter Yared as an alternative to buying dedicated hardware for a private local model. Yared is Founder and CEO of InCountry , the data residency company behind AgentCloak, which raised $10 million in funding alongside the launch of AgentCloak for AI agent data protection . The paid AgentCloak Server tier extends the same approach to teams handling patient records and other regulated data across borders.
Comments (12)
Before I take this to legal, is there a SOC2 Type II report and does the cloaking happen on-prem or in your cloud? Data residency answers determine whether procurement even opens the deck.
hot take: the real moat here isn't the cloaking, it's convincing employees to stop pasting the customer list into ChatGPT at 11pm.
Distribution question: are you selling this to CISOs or sneaking in bottom-up via a browser extension? Very different motions and your landing page hedges both.
Per-token pricing, seat-based, or flat platform fee? Asking because 'inspects every AI prompt in the org' sounds like a line item that grows faster than headcount.
Reminds me of what a couple portfolio cos tried to build in-house and shelved. If you nail the enterprise wrapper this is a very easy intro conversation.
So it's a bouncer for your prompts checking IDs at the door. I respect it.
Is there a proxy endpoint I can point my SDK at, or does every model need a bespoke integration? Also curious about rate limits when the cloaking layer is in the hot path.
The tweet buries the actual demo. You had one job: show the before/after prompt side by side in the first 3 seconds.
Been circling this problem from the compliance angle at my shop. Curious how you handle re-identification on the response side without leaking the mapping table.
That tweet cuts off mid-sentence at 'actual https://t.co/...' which is either the worst luck or the best engagement hack I've seen this week.
Peter, off the record or on, are the Chinese AI services actually customers already or is that positioning? Happy to keep it background.
Cloaking data before sending to third parties feels like it wants a verifiable on-chain attestation layer. Just saying, the primitive is right there.