How Tripwire works
You plant a decoy credential where an intruder would look. It grants access to nothing of yours. The moment someone uses it, you get an alert that tells you who, from where, and what they tried. This page covers the detection model, the alert you get, the safety model, and how to place your first tripwire.
The detection model
Tripwire is deterministic detection, not probabilistic. A normal security tool guesses whether activity looks malicious. A tripwire has no legitimate use: nobody should ever authenticate with the decoy credential you planted, so any use is a real signal, with near-zero false positives.
The alert fires on use, not on exposure. Reading the file, viewing the key, or copying it does nothing. The alert fires when the credential is presented to the real service, for example an AWS sts:GetCallerIdentity call with the decoy key. That is the moment an intruder has both found and acted on your secret.
What you can plant
- AWS access key: fires on any authenticated AWS API call (captures source IP, user-agent, and the action).
- GitHub token: fires when the token is used against your organization’s GitHub.
- Anthropic API key: fires when the decoy key is used against the API.
- Planted service credentials: web-login, browser-session cookie, PostgreSQL URL, and Kubernetes kubeconfig decoys that fire when presented to Tripwire’s listener.
The downloadable kit ships a ready-to-place decoy (an AWS key inside a believable project). For GitHub, Anthropic, and the service credentials, use the CLI, the Claude Code plugin, or the web console to mint exactly the canary you want.
The alert
When a decoy is used, Tripwire emails you a single high-confidence alert with the full event:
- What: the action taken (e.g.
sts:GetCallerIdentity). - From where: source IP, ASN, and user-agent.
- When: the timestamp of the use.
- Which: the canary that fired and the memo you gave it, plus a console link.
Latency. Planted service-credential canaries fire in about a second. Cloud-provider canaries (AWS, GitHub, Anthropic) depend on the provider’s own audit stream and typically alert within a few minutes, occasionally longer. If you want instant confirmation that your pipe works without waiting for a real intrusion, contact us for a test fire.
Is it safe to plant these? (the isolation model)
Yes. Every decoy is minted in a quarantined account that Tripwire controls, isolated from your systems. The credential is real, so an attacker trusts it, but it can only ever reach Tripwire’s own bait resources, never anything of yours. Nothing runs on your machine, there is no agent, and Tripwire is never given a role or token in your cloud. You receive an inert file; you grant no access. See our Security page for the full posture.
Getting started
Three ways in, pick whichever fits how you work:
Download a kit (no account). Enter an email on the home page, download the kit, and place the decoy file where a snoop would look (an old repo, a shared drive, the docs you feed a coding agent). Done.
Claude Code plugin. In Claude Code:
/plugin marketplace add tachyon-oss/tripwire /plugin install tripwire@tripwire /tripwire:init
The plugin runs its CLI via uv (install uv if you don’t have it).
CLI (terminal-first):
npm install -g @tachyonhq/tripwire # or: uv tool install tripwire-cli tripwire canary create aws.access_key # place the returned key, then any use of it emails you
There is no separate sign-up step: the first command that needs an account signs you in, passwordlessly, with a six-digit code emailed to you. Canaries and fired events are visible in the web console and via tripwire status.