Local-first · agent-agnostic

Let agents read your code,
not your secrets.

Fida installs local protection for AI coding agents, verifies it with a synthetic credential, and scans repository risk. Secret values are redacted before reaching the model.

protected

DEMO_CREDENTIAL=

SAFE VIEW — detected value redacted

useful structure returned · the secret never reached the model

Get startedView on GitHub

$curl -fsSL https://raw.githubusercontent.com/ajipurn/fida/main/install.sh | sh

Coding agents read everything you point them at — including the .env you forgot was there.

install · verify · scan

Three commands. Honest coverage.

Install protection for detected agents, verify the real read and shell paths, then see whether a raw secret value can still reach a model.

01 / protect

Install agent protection

One command detects every supported agent, installs its redacting gateway and steering, and runs a synthetic-secret self-test. Protection is global — one install guards every repository.

fida

installing

$fida

◇ detecting agents…

✓ Codex enforced

✓ Claude Code enforced

✓ Cursor best_effort

✓ Windsurf best_effort

◇ synthetic-secret self-test … passed

✓ protection installed · global

02 / verify

Know your coverage

See enforced, best-effort, or incomplete protection per agent — alongside the count of secrets Fida has already protected in this repository. No guessing what is actually guarded.

fida status

coverage

$fida status

Codex enforced

Claude Code enforced

Cursor best_effort

Copilot best_effort

────────────────────────────

12 secrets protected · this repo

03 / scan

Find raw-secret risk

Scan tracked and sensitive files, then read the answer that matters: raw_secret_exposure — whether an unredacted value could ever reach a model. It never prints a secret value, length, or fragment.

fida scan

scanning

$fida scan

scanning tracked + sensitive files…

.env secret detected · git-tracked

risk medium

protection enforced

raw_secret_exposure false

→ no unredacted value reaches a model

Codex✳
Claude Code✳
Cursor✳
OpenCode✳
Windsurf✳
Copilot✳
Antigravity✳

Protection you can verify.

One command installs supported integrations, runs a synthetic-secret self-test, and scans your repository. Fida reports enforced and best-effort coverage honestly.