It is a local proxy that swaps secrets for coding agents.

Decoyrail: endpoint firewall for AI agents
Questions about Decoyrail: endpoint firewall for AI agents
What is Decoyrail: endpoint firewall for AI agents?
Decoyrail is a cybersecurity tool that acts as an endpoint firewall for AI coding agents. It runs as a single binary that places agents like Claude Code or Codex CLI behind a local proxy. The system replaces real credentials with decoys in the agent's environment, intercepting outgoing requests to only swap in actual secrets for pre-approved destinations while blocking and logging other attempts.
Who exactly does this tool suit and who might it not be for?
The product is designed for users who want to secure their coding agents against credential leaks. It automatically detects credential-looking environment variables and substitutes decoy values, keeping real secrets encrypted on disk and accessible only within the proxy. This setup ensures that if an agent is tricked into leaking a key, the leaked key works nowhere because it is a decoy.
How can a new user get started or order the software?
Getting started is quick and requires no complex vault setup for common cases. Users can run the agent behind Decoyrail with a single command, such as 'decoyrail run -- claude', which installs a default-deny egress policy and replaces credential-looking env vars with decoys automatically. The whole process can be set up in about 5 minutes.
What are the delivery methods and licensing terms?
The software is delivered as a single binary that functions as an endpoint firewall. It is available with installation instructions specifically for macOS. The core of the product is licensed under FSL-1.1-ALv2, which becomes Apache-2.0 after two years, and the source code lives on GitHub.
How does the proxy mechanism work in detail?
The proxy intercepts outgoing requests and only swaps in the actual secret for pre-approved destinations, blocking and logging any other attempts. It never follows a redirect, so a secret cannot be bounced somewhere the policy never looked at. This narrow swap ensures that the real secret goes out only when the host, path, method, and location all match what was bound.
What specific credential types does the system detect and replace?
Decoyrail automatically detects credential-looking environment variables such as AWS_SECRET_ACCESS_KEY and GITHUB_TOKEN. It hands the agent decoys automatically, ensuring that the agent, its logs, its prompts, and its crash dumps can only ever contain fakes that look real, like sk-ant-… or ghp_…. Stock SDKs accept these fakes without any special-casing.
How are TLS and certificate management handled?
The system uses a CA minted on the user's device to handle TLS interception carefully. It creates per-host certificates and ensures a fresh, fully verified connection upstream. Enterprise internal CAs are additive and never act as a bypass, maintaining security while allowing necessary integrations.
What features ensure performance and log integrity?
Token streams pass through untouched to keep streaming fast, while bounded responses are scanned for echoed real secrets. The audit log is append-only and hash-chained, meaning decoyrail log --verify catches any edits to ensure the log you can trust remains intact.
How does the system alert users to unauthorized credential usage?
Every decoy acts as a honeytoken that tells on the thief if it shows up outside its binding. Even if the decoy is base64, hex, or percent-encoded, the request is blocked and the alarm is logged the moment anyone tries to use it outside the approved scope.
What does the default configuration include for new users?
The starter pack covers what coding agents need, including a default-deny egress policy where nothing leaves unless a rule allows it. The rule that allows a destination also specifies which secrets it releases, providing a comprehensive security baseline out of the box.
How can visitors contact the business or find more information?
The business provides a contact email at [email protected] for inquiries. The source code is available on GitHub, and the website domain is decoyrail.com for further information and updates.
What is the official website of Decoyrail: endpoint firewall for AI agents?
The official website of Decoyrail: endpoint firewall for AI agents is decoyrail.com.
How is Decoyrail: endpoint firewall for AI agents rated on Trustburn?
On Trustburn, Decoyrail: endpoint firewall for AI agents has a rating of 4.3 out of 5 based on 3 reviews.
Post your question publicly — other visitors, and whoever manages this page, can answer. It will be visible on this page.
Want to hear about this company again? Follow it and we’ll email you when a new review appears.
Based on reviews, listing data and the company's website. Verify anything important with the company directly.
The installation process seems straightforward, requiring just a few commands. Having default-deny egress and an audit log ready immediately is convenient. The proxy handles credential swapping automatically, which reduces manual configuration overhead. It appears to be a practical solution for developers who want to secure their agents without complex infrastructure changes.
Decoyrail offers a clever approach to securing coding agents. By using decoy credentials, real secrets never enter the agent’s environment. The local proxy swaps in real secrets only for approved destinations, blocking everything else. It’s a single binary that runs behind your agent, providing default-deny egress and an audit log. This setup feels robust for preventing accidental leaks.
Questions & Answers
No questions yet — be the first to ask.
Sign in to post
Your text is saved — you'll come right back to post.
Don't have an account? Create one
News & Press
No news yet. This is where the business shares its own updates and press.
Photos
No photos yet. Photos shared by reviewers will show up here.
About Decoyrail: endpoint firewall for AI agents
Decoyrail is a cybersecurity tool that functions as an endpoint firewall for AI coding agents. It is delivered as a single binary that runs agents such as Claude Code or Codex CLI behind a local proxy, replacing real credentials with decoys in the agent's environment. The proxy intercepts outgoing requests and only swaps in the actual secret for pre-approved destinations, blocking and logging any other attempts to use those credentials. Decoyrail automatically detects credential-looking environment variables and substitutes decoy values, while the real secrets remain encrypted on disk and accessible only within the proxy. The product is available with installation instructions for macOS, accompanied by documentation, a pricing page, a roadmap, and a GitHub repository.
- Website
- decoyrail.com
- [email protected]
- Categories
- Computer & network security