Edgehound

io.github.EnergixePublished by GitHub account Energixev1.1.0Updated Oct 24, 2026

Writes adversarial test suites for AI-built code. Your agent's test engineer.

VerifiedStreamable HTTPWeb executableDeveloper ToolsAI & ML

Overview

AI-generated overview

Hosted MCP server that writes adversarial, runnable test suites for AI-built code and predicts which bugs they will expose.

What you can do
  • Generate a full adversarial test suite for a given source file, covering boundary values, unicode, timezone and month-end math, money rounding, double submits, injection strings and pagination edges.
  • Predict, before running anything, which bugs the suite will expose, with the code evidence for each.
  • Produce output for JavaScript/TypeScript (Jest, Vitest, node:test), Python (pytest), Go, PHP (PHPUnit), Ruby (RSpec) and Rust (cargo test), with syntax-validated output and auto-repair.
  • Work from multi-file context so the generated tests fit the surrounding code.
When to use it
Useful when an agent writes or changes code that accepts input, handles money, dates or state, and you want a hostile test suite plus a predicted-bug report without writing tests by hand. Also handy for adding coverage to an existing helper module in a supported language.
Requirements
A remote MCP endpoint added to the client config; no local package or build. Sign-in through MCPize OAuth on first use. An MCP-capable client such as Claude Code, Cursor, opencode or VS Code.
Before you install
The generation engine is a hosted third-party service, so the source files you point it at are sent off your machine. Sign-in is required and tied to an MCPize account. Generated tests are meant to be saved and run locally; review them before trusting or committing them.

Installation

In SourceWeft

  1. Open Edgehound in the dashboard and add it to a workspace.
  2. Connect the required credentials. They are stored privately and never shown on public pages.
  3. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Use with an agent

Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.

Read https://sourceweft.com/mcp/io.github.Energixe%2Fedgehound-mcp/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.

Read the install guide the agent follows

Other MCP clients

Add this to your client's mcpServers config. The client completes OAuth sign-in on first connect.

{
  "mcpServers": {
    "edgehound": {
      "type": "http",
      "url": "https://gauntlet.mcpize.run/mcp"
    }
  }
}

README

Edgehound

An MCP server that writes test suites designed to break your code.

This repo is the public front door — install guide, examples, and issue tracking. The generation engine runs as a hosted service, so there is nothing here to build or self-host.

What it does

Point your AI agent at a file, and Edgehound writes a complete, runnable test suite that attacks it: boundary values, unicode and emoji, timezone and month-end math, float money rounding, double submits, injection strings, pagination edges. It also predicts — before you run anything — which bugs the suite will expose, with the exact code evidence for each.

Works with Claude Code, Cursor, opencode, VS Code, and any MCP client.

Install

Add the server to your MCP config:

json
{  "mcpServers": {    "faultline": {      "url": "https://gauntlet.mcpize.run/mcp"    }  }}

On first use your client will ask you to sign in to MCPize. After that:

"write tests for src/pricing.ts"

and your agent does the rest.

Make it automatic

Paste this into your project's rules file (CLAUDE.md, AGENTS.md, or .cursor/rules), and your agent will reach for Edgehound every time it writes or changes input-handling code — no prompting needed:

Whenever you create or modify code that accepts input, touches money, handlesdates, or manages state — call the Edgehound tool (generate_test_suite) beforemarking the task complete. Save the returned suite next to the file, run it,and fix any failing test by fixing the code, not by weakening the test.Skip only for pure UI markup, copy changes, and configuration.

What a run looks like

examples/user-utils.js has four innocent-looking helpers. A real generation produced 60 tests — 55 passed, 5 failed. Every failure was a real bug in the sample:

  • parseDuration('1h30m') returns 3600 — it silently drops the minutes
  • monthlyPrice leaks float drift on non-divisible cents
  • half-value money rounding is unhandled
  • negative input produces a negative price
  • truncate blows up when handed a number instead of a string

Try it:

bash
node --test examples/user-utils.test.js

Languages today: JavaScript/TypeScript (Jest, Vitest, node:test), Python (pytest), Go, PHP (PHPUnit), Ruby (RSpec), Rust (cargo test).

Issues

Bugs, weird output, missing languages — open an issue here. We read everything and ship fixes fast. Want Go/Rust/PHP improvements or a language we don't cover yet (Java, C#, Swift)? Say so in an issue; demand decides what gets added next.

Status

v1.1 — predicted-bug reports, syntax-validated output with auto-repair, multi-file context, 8 test frameworks.

Source: README.md at commit 4c04a80

Tools

0
This server asks you to sign in, so SourceWeft could not read its tools without an account and has not indexed them. What a connected account sees can depend on that account and its permissions.

Version history

1
  1. v1.1.0LatestSep 17, 2026

Score

85% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 daysRequired25 pts
  • Auth declaredRequired10 pts
  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts