Heldkeys

Heldkeys is a CLI tool that lists credentials an AI agent can access on a machine or in CI without printing secrets.

Heldkeys identifies which credentials an AI agent can use if a prompt injection hijacks it. It scans the current process environment, cloud CLI logins, AI tool logins, developer platform CLIs, package and git tokens, git config, project files, SSH keys, MCP server configs, and host exposures like Docker sockets and metadata endpoints. The tool runs locally on Linux and macOS as a single self-contained binary. It operates in a read-only manner, performs no telemetry, and outputs fingerprints instead of printing actual secrets. It is built for developers who run coding agents locally or in CI environments and want to secure their exposed credentials.

Heldkeys at a glance

Best for: Software engineer
Runs on: Linux, macOS
Pricing: Free
Released: October 2026
Category: AI coding tool
Website: github.com/hugomn/heldkeys

Key features

Scans environment variables and cloud CLI logins
Finds AI tool and developer platform credentials
Inspects package manager and git credentials
Detects SSH keys and MCP server tokens
Checks sandbox and CI job exposures
Outputs token fingerprints instead of secrets

Use it for

Audit local machine credentials accessible to AI coding agents
Inspect GitHub Actions and GitLab CI jobs for exposed secrets
Check sandbox and container environments for security risks

Alternatives to Heldkeys

Weir — It detects exact key values in agent logs and blocks them in coding sessions.
Hlock — Pins entire AI agent setups into a portable file to share and run anywhere.
Varlock — It secures environment variables for both humans and AI agents.
OpenHands — Runs autonomous agents locally, in the cloud, or inside your VPC with complete control.
jevgrep — Reduces coding agent costs by providing targeted, relevant repository context.
LeakLock — It scans live sites to find keys missing from git history and generates AI-builder fix prompts.
Portkey — Unified API access with observability, guardrails, and caching for LLM applications.
mcpgawk — Pins each tool to what you approved and blocks calls if a server changes.
Coucou — It provides open-source, real-time agent monitoring and approvals directly from the screen notch.
OpenClaude — Connects one terminal interface to over 200 different AI models.
Stroq — Scans recorded sessions and guards future actions with local hooks.
Agent Sandbox — Provides a secure and isolated execution layer for autonomous AI agents.

All Heldkeys alternatives