Skip to content

Hide private details in one call.

Point your OpenAI or Anthropic client at Secure AI. Personal details come out before every request and go back into every answer. 50,000 requests a month free.

Get a free API key
  • Redact

    Real details out. Stand-ins in.

    Redact

    Real details out. Stand-ins in.

    No session and no SDK. Send text and get it back with names, emails, phone numbers, addresses, cards and account numbers swapped for believable stand-ins, plus a map of what was swapped.

    Read the docs
  • Restore

    And back again.

    Restore

    And back again.

    Send the model's reply and the map to /v1/restore, and the real names, numbers and addresses go back in. The map stays with you; Secure AI keeps no copy.

  • Drop-in Proxy

    Already using an OpenAI client? Change one line.

    Drop-in Proxy

    Already using an OpenAI client? Change one line.

    Point your existing client at /v1/proxy. Details are taken out on the way through and put back in the reply, using your own model key. Tool calls work too: results are hidden like everything else, and each call comes back with the real values, checked against your rules first.

    Quick start
  • MCP

    Listed in the official MCP registry.

    MCP

    Listed in the official MCP registry.

    Agents that speak MCP reach the same controls as tools at /mcp: hide details, put them back, and check an action before it runs. Building your own MCP server? Start from create-secure-mcp.

    Build a secure MCP server
  • n8n

    A Secure AI step for any workflow.

    n8n

    A Secure AI step for any workflow.

    Install the Secure AI node, or point n8n's OpenAI credential at Secure AI. Hide details before any AI step and put them back after.

    Using n8n
  • Zapier

    Coming soon to Zapier.

    Zapier

    Coming soon to Zapier.

    Two actions to put either side of an AI step in a Zap: Hide Personal Details, then Put Details Back. Available by invitation while it waits for Zapier's public listing.

  • SDKs

    TypeScript, Python, Swift and Kotlin.

    SDKs

    TypeScript, Python, Swift and Kotlin.

    All four are MIT licensed, with the source on GitHub. npm install @secure-ai/guard, or pip install secure-ai-guard.

    GitHub

Govern every move your agents make.

Quick start
  • Inspect

    Every action, judged before it runs.

    Inspect

    Every action, judged before it runs.

    /v1/inspect judges whether an action may happen at all. Send the tool call your agent was about to make; the decision comes from rules held on your account, not from anything in the request. Approve waits for a person.

  • Guard

    Wrap a function. It runs safely, or not at all.

    Guard

    Wrap a function. It runs safely, or not at all.

    The SDK wraps a function your agent already calls and runs it with the rewritten arguments, so protection does not depend on anybody remembering to check a decision. A refusal throws.

  • What It Finds

    Finds what's personal. Even without a label.

    What It Finds

    Finds what's personal. Even without a label.

    A number with no label, like “payment from 10972500”, is read in its sentence and taken out when it is an account, an ID or a place. You can add any term of your own.

  • Rules

    Start from PCI, HIPAA or GDPR.

    Rules

    Start from PCI, HIPAA or GDPR.

    Rules per kind of data, per tool and per direction, starting from a template rather than an empty file. Allow your own domain once, for every agent.

  • Approvals

    A person decides.

    Approvals

    A person decides.

    An action a rule holds waits in Approvals. If nobody answers in time, the action is denied.

  • Audit Log

    Everything your agents did. None of the details.

    Audit Log

    Everything your agents did. None of the details.

    Kinds and locations of what was found, the decision and the tool, never the values themselves.

  • Gateway

    Every request, checked on the way out.

    Gateway

    Every request, checked on the way out.

    Point an HTTP client at /v1/gateway/<url> and every request gets the same inspection on its way out.

Start free. Grow with us.

Start with one line.

Your model, your key, your bill. Secure AI keeps the private details out of it.

50,000 requests a month free. No card needed.

Secure AI

Use any AI. Stay anonymous.

© 2026 Secure Artificial Intelligence, Inc.Private details never reach the AI.
Secure AI
Developers · Secure AI