Skip to main content
fhenix-toolkit is a Claude Code plugin maintained by Fhenix. It teaches the assistant the rules that make confidential contracts work. Those include FHE.select instead of branching on an ebool, the access control calls a stored value needs, which decryption flow fits a feature, and how to test against the mocks. Use it when you build with FHE.sol or @cofhe/sdk in Claude Code. The plugin carries concepts, decision trees, and a catalog of known mistakes. Function signatures, SDK methods, error codes, and deployed addresses are looked up live from the public Fhenix repositories when they are needed. The guidance therefore tracks new releases instead of freezing at one API version.

Install

1

Add the marketplace and install the plugin

In a Claude Code terminal session:
In the Claude Code desktop app, open Customize, then Personal Plugins, add the marketplace FhenixProtocol/fhenix-toolkit, sync, and add the Fhenix Toolkit plugin.
2

Ask for something confidential

Each skill activates on its own when the context matches, so there is nothing to invoke by hand. Try:
  • “Help me write a confidential ERC-20.”
  • “Audit this FHE.sol contract.”
  • “Write Foundry tests for this encrypted auction.”
When the plugin is active, Claude refuses to write if (FHE.gt(a, b)) and routes the comparison through FHE.select. That is the quickest way to confirm it loaded.

What the plugin does

The fhe-reviewer subagent backs fhenix-review for large diffs and pre-launch audits. It loads the full catalog into a fresh context, walks the code function by function, and returns a prioritized report.

Using other assistants

The skills are Markdown files under plugins/fhenix-toolkit/skills/ in the repository. If you work in Cursor, ChatGPT, or another assistant, load the SKILL.md for the area you are working in as context, along with its references/ folder. Every page on this site also carries a menu to open it in ChatGPT, Claude, or Cursor. That is the fastest way to give an assistant the current documentation for one topic. For prompting techniques that work with any assistant, see build with AI.

Status and feedback

The plugin is public and pre-1.0. Coverage gaps are tracked in the repository’s docs/known-flaws.md. Report problems or suggest additions by opening an issue on the fhenix-toolkit repository.