Does Sheathe work with Cursor?

Yes, Sheathe works seamlessly with Cursor IDE. It intercepts requests before they reach OpenAI, masking sensitive data while maintaining full Cursor functionality including chat, autocomplete, and codebase analysis.

Is Sheathe compatible with Claude Code?

Yes, Sheathe fully supports Claude Code CLI. It masks sensitive information before it reaches Anthropic's servers while preserving all Claude Code features and functionality.

How much latency does Sheathe add?

Sheathe adds only ~90ms of latency to each request. Your AI coding assistants remain fast and responsive.

Does Sheathe run on my infrastructure?

Yes, Sheathe runs completely on your infrastructure. Your sensitive data never leaves your premises and you maintain full control over the privacy layer.

Let AI see your code.
Not sensitive data.

Mask confidential information locally in your code
before Cursor or Claude Code sends it to OpenAI or Anthropic

Sheathe sensitive data

Your Code

db = connect('prod-db')
users = db.query('payments',
  {api_key: 'sk-live-123'})
fetch('api.internal.com')

Exposed data

sheathe

Fully local proxy

What AI sees

db = connect('[DB_1]')
users = db.query('[TABLE_1]',
  {api_key: '[API_KEY_1]'})
fetch('[INTERNAL_URL_1]')

Sheathed data

99.7% of your data
stays private

And 100% coverage for your defined Regex patterns

0.3%

False negatives

100%

On your premises

~90ms

Added latency

Take control of
data masking

20+ data types covered by default.
Easily add new data types or remove existing ones through JSON configuration.

API keys
Database passwords
OAuth tokens
JWT secrets
AWS credentials
Service account keys
Webhook URLs
Connection strings
Email addresses
Phone numbers
Credit card numbers
SSNs
Personal information (PII)
User IDs
Internal URLs
Company names
Project codenames
Employee names
Server hostnames
IP addresses
Docker registry URLs
Internal endpoints
Environment variables
Anything else you configure :]

API keys
Database passwords
OAuth tokens
JWT secrets
AWS credentials
Service account keys
Webhook URLs
Connection strings
Email addresses
Phone numbers
Credit card numbers
SSNs
Personal information (PII)
User IDs
Internal URLs
Company names
Project codenames
Employee names
Server hostnames
IP addresses
Docker registry URLs
Internal endpoints
Environment variables
Anything else you configure :)

Host our privacy layer
on your premises

Download Sheathe

Deploy via a Bash one-liner

Configure your sensitive data types

Let it index your codebase, docs and tickets

Done - Sheathe is ready to process requests

Then update
the LLM endpoint
in your AI assistant

Cursor

Claude Code

Cline

Soon

| Plan, search, build anything. Ask anything about your code

More integrations are coming soon

We detect and redact sensitive entities (like API keys, passwords, names, emails, and more) using a combination of regular expressions, pattern-matching, and our own LLM classifier that runs on your infrastructure for intelligent context-aware detection.

Redacted tokens are replaced with deterministic placeholders (e.g., [NAME_1]) that can later be accurately restored in the OpenAI API response.

No. We preserve the semantic structure of the prompt. Entities are replaced with realistic placeholders to maintain context. E.g.:

user@email.com → EMAIL_1

LLMs still understand what's being asked, and you get usable, high-quality responses.

We don't see or store any of your inputs, outputs, or tokens. Everything runs inside your environment.

The only data we store in our database comes to us via this website. It is your user account data when you sign up (email, password, purchases, notification settings).

But Sheathe as a tool doesn't send data anywhere—it processes it locally on your premises.

When Sheathe has indexed your codebase, it takes around 50-200ms to mask or unmask the request/response.

Indexing time depends on your codebase. It takes around 5 minutes to fully scan a popular library like Pandas.

Let AI see your code.
Not sensitive data.

99.7% of your data
stays private

Take control of
data masking

Host our privacy layer
on your premises

Then update
the LLM endpoint
in your AI assistant

Cast privacy shield
on your team

Pricing

Frequently Asked Questions

Qwen or DeepSeek
isn't for everyone

Let AI see your code. Not sensitive data.

99.7% of your data stays private

Take control of data masking

Host our privacy layer on your premises

Then update the LLM endpoint in your AI assistant

Cast privacy shield on your team

Pricing

Frequently Asked Questions

How does it work?

How do you process requests from us?

Will it break LLM understanding or outputs?

Will it scan .env?

Do you log or store our data?

What's the processing latency?

Can I control what gets redacted?

Qwen or DeepSeek isn't for everyone

Let AI see your code.
Not sensitive data.

99.7% of your data
stays private

Take control of
data masking

Host our privacy layer
on your premises

Then update
the LLM endpoint
in your AI assistant

Cast privacy shield
on your team

Qwen or DeepSeek
isn't for everyone