Back to browse
GitHub Repository

fakekey - never expose real API keys in the agent environment.

16 starsRust

Fakekey – never expose real API keys in the agent environment

by tomsun28·Mar 31, 2026·2 points·0 comments

AI Analysis

●●SolidNiche GemBig Brain

MITM proxy swaps fake keys for real ones so agents never see credentials.

Strengths
  • Rust implementation ensures low latency overhead during header substitution for all requests.
  • Prevents accidental key leakage via prompt injection or context window exposure risks.
  • Cross-platform binaries with easy Homebrew and Cargo installation options available now.
Weaknesses
  • Requires installing a root CA certificate, which raises significant trust concerns.
  • No Windows GUI for managing keys, CLI-only configuration currently available today.
Category
Target Audience

AI agent developers, Security engineers

Similar To

HashiCorp Vault · Doppler · mitmproxy

Post Description

With the widespread adoption of AI Agents, configuring various service API Tokens directly in environment variables has become common practice. Your api_key will be inserted into context and known by model service providers, known by the lobsters you trust, perhaps captured and read by some skill, and even more likely to be directly learned by strangers asking your lobster. There are too many leak cases, I cannot trust to bind my credit card-linked api_key directly exposed to any Agent and local environment variables, so FakeKey was created, the safest measure is to never expose the real api_key.

FakeKey is a high-performance API key proxy program developed in Rust. Through intelligent proxy technology, it can automatically replace fake keys with real keys in HTTP headers and URLs without exposing real credentials, while maintaining complete HTTP API compatibility and performance.

Similar Projects

Security●●Solid

OneCLI – Vault for AI Agents in Rust

Agents never see real keys, but Vault already does secret injection.

Solve My ProblemSlick
guyb3
161523mo ago
Security●●●Banger

Phantom – Let AI use your API keys without leaking them

Proxy tokens worthless if leaked, real keys never enter LLM context windows.

Big BrainSolve My ProblemDark Horse
masonwyatt23
202mo ago
Security●●Solid

AVP – an agent can't leak a secret it never had

Agents can't leak secrets they never had—placeholder injection at the wire.

Big BrainSolve My Problem
radku
314d ago