Skip to content
Maskflare
Talk to usStart free trial
Menu

For your agents · Maskflare Coding Agents

Keep Secrets and Customer Data Out of Claude Code, Cursor and Codex

Developers paste stack traces, .env files and customer records into coding agents every day. Maskflare hooks stop prompts that carry secrets and mask what the agent's tools read and write, before it reaches the model.
Start free trial Talk to us
28-day free trial or 10,000 requests, whichever comes first. No card needed.

What the developer pastes

Why does this fail?

OPENAI_API_KEY=sk-proj-7Hq2…Xv9

DATABASE_URL=postgres://app:S3cr3t!pw@db

customer: ada@example.com

Maskflare

What the coding agent sends

Why does this fail?

OPENAI_API_KEY=mf_mask_c03a

DATABASE_URL=postgres://app:mf_mask_88b1@db

customer: mf_mask_5e1d

Claude Code · prompt hook · secret blocked before the model

574

Built-in detectors for personal data and secrets

47

Country packs, checksum-validated where the format has one

9

AI providers behind one gateway, plus the desktop agent and proxy

Sound familiar?

Why the usual answers fall short

“We have secret scanning in CI.”

It runs after the commit. The agent sent the key to the model an hour earlier.

“Our devs know better.”

They do, until they paste a failing request with the Authorization header still in it.

“We'll just block the agent.”

Then you lose the productivity you bought it for. Mask the secret, keep the agent.

A coding agent reads files, runs shell commands and fetches URLs on its own. One cat of a .env file or a database dump and the credentials are in the next request to the model.

Secret scanners in CI catch keys after they're committed. By then the same key has already been sent to an AI provider, often more than once.

How Maskflare Coding Agents works

The maskflare-hook command plugs into each coding agent's hook system. In Claude Code it blocks a prompt that contains a match and masks the input and output of Bash, WebFetch, Write, Edit and MCP tools. Cursor and Codex CLI get the protections their hook APIs allow. Detection uses the same 232 credential formats and personal-data detectors as the rest of Maskflare.

  1. 01

    Install the SDK

    pip install maskflare on the developer machine or in your devcontainer image.

  2. 02

    Install the hook

    Run maskflare-hook install for Claude Code, Cursor or Codex CLI. It writes the agent's hook configuration.

  3. 03

    Work as normal

    Prompts with secrets are stopped with a clear message; tool input and output is masked before the model sees it.

Benefits

Stops secrets at the prompt

A prompt that contains an API key, token or private key is blocked before it's sent.

Masks tool traffic

Shell output, fetched pages and file edits are masked before they go back into the model's context.

232 credential formats

Cloud keys, AI provider keys, database URLs, private keys, OAuth tokens and more.

Same policy

Rules come from your Maskflare environment, so security sets them once for everyone.

Capabilities

Claude Code

Prompt blocking plus masking of Bash, WebFetch, Write, Edit and mcp__* tool input and output.

Cursor and Codex CLI

Prompt and tool protections within the limits of each agent's hook API.

MCP

MCP servers the agent uses can also be routed through the gateway's MCP route.

Set it up

Install for Claude Code
pip install maskflare
maskflare-hook install claude-code
Install for Cursor or Codex CLI
maskflare-hook install cursor
maskflare-hook install codex

What it does not do

We'd rather you hear it from us.

Explore related capabilities

See how Maskflare applies the same masking rules to related AI and data-protection workflows.

Maskflare Detection EngineMaskflare AI GatewayMaskflare Forward ProxyMaskflare Exact Data MatchMaskflare File SecurityMaskflare AI App Discovery & ControlMaskflare User CoachingMaskflare WorkforceMaskflare Browser ProtectionMaskflare Agent & MCP SecurityMaskflare SDKs & APIMaskflare Admin Console

Related guides and explainers

Frequently asked questions

Which coding agents are supported?

Claude Code, Cursor and Codex CLI. Claude Code has the widest coverage because its hook API exposes the most.

Does it slow the agent down?

Masking adds a few milliseconds per call with the default rules, which you won't notice next to model latency.

What does the developer see when a prompt is blocked?

A short message naming the rule that matched, never the secret itself, so they can remove it and try again.

Can we enforce it for every developer?

Install the hook in your devcontainer or base image and manage the rules centrally in the Maskflare console.

See it on your own data

Try Maskflare free.
Bring your hardest prompts.

28-day free trial or 10,000 requests, whichever comes first. Point one app at the gateway or enroll a few laptops,
and watch what gets masked. Or talk to us first.

Start free trial Talk to us