Earnie

Your code, checked before anyone else sees it.

For developers. Earnie checks your work where you already do it: in your coding agent through MCP, in your terminal, at pre-commit, and at merge on GitHub and GitLab.

A developer with headphones working at a laptop, with a terminal showing an Earnie pre-commit check passing.
  • Claude Code
  • Cursor
  • Codex

Four ways in, one policy

Earnie meets your change wherever it is: in your coding agent’s session, in your terminal, at the commit and at the pull request. The same project policy runs at every one, so whichever you reach first gives the same answer.

MCP: Earnie in your coding agent

Earnie is there when AI is looking for guidance.

Your agent pulls in packages, snippets and dependencies faster than anyone can review them. Connected to Earnie’s MCP server, it reads the project’s policies before it writes and sends its changes to earnie_review_code before it shows them to you: a new dependency, code copied from elsewhere, cryptography, a call to an AI SDK.

Works with Claude Code, Cursor, Codex and your IDE.

How an agent session works
Add a revenue chart to checkout
I’ll use fast-charts. Checking it with Earnie first.
earnie_review_code 2 files
gate block fast-charts@4.2.1 is AGPL-3.0
fix use an MIT- or Apache-2.0-licensed library
That licence isn’t allowed in checkout-service, so I switched to an MIT-licensed library. Reviewed again: pass.

An example session. The package and project are made up.

Earnie CLI

Free. Local. Fast.

Scan a directory, check a dependency, verify a snippet, from your terminal, before anyone else sees it. Detection is free.

earnie check tells you what some files bring in and whether they break a project policy, without touching the project. earnie model scan uploads an AI model file, such as a .safetensors or .gguf, and tells you what it really is.

Install the CLI ↗
$ earnie check src/charts/ --project checkout-service
gate block
fast-charts@4.2.1 AGPL-3.0 no-copyleft
$ earnie scan path .
↳ 412 files. 12 components. 0 violations.

Example output, shortened.

Pre-commit hooks

Catch it before it’s committed.

Earnie fires from your pre-commit hook and flags the violation in your console, before the code leaves your machine.

Run earnie hook install once in the repository. The hook checks exactly what you’ve staged, stops the commit on a block, and links to the full result in the web app.

$ git commit -m "add revenue chart"
project checkout-service
files 3 submitted, 3 included
verdict block
src/charts/revenue.ts:3 fast-charts@4.2.1 AGPL-3.0
commit stopped

Example output, shortened.

Merge-time checks: GitHub and GitLab

The gate that documents itself.

Every pull request checked against project policy. Pass, warn, or block, and every decision lands in the record.

Only the findings a pull request introduces decide its check. Each one gets an inline comment on the line that added it, with a suggested fix. On GitHub, make Earnie a required status check and a block stops the merge. On GitLab the check is advisory for now.

Add revenue chart #312
EarnieBlocked by a finding this change introduced
src/charts/revenue.ts
+ import { Chart } from "fast-charts";
Earnie fast-charts@4.2.1 is AGPL-3.0, which the no-copyleft policy blocks. Suggested fix: use an MIT- or Apache-2.0-licensed alternative.

An example pull request. The package and project are made up.

Getting started

Install the CLI, connect the MCP, add the Action.

Read the docs ↗

Checked before the code lands.

Request a demo
Earnie for developers: CLI, MCP and CI/CD
Earnie

Your code, checked before anyone else sees it.

For developers. Earnie checks your work where you already do it: in your coding agent through MCP, in your terminal, at pre-commit, and at merge on GitHub and GitLab.

A developer with headphones working at a laptop, with a terminal showing an Earnie pre-commit check passing.
  • Claude Code
  • Cursor
  • Codex

Four ways in, one policy

Earnie meets your change wherever it is: in your coding agent’s session, in your terminal, at the commit and at the pull request. The same project policy runs at every one, so whichever you reach first gives the same answer.

MCP: Earnie in your coding agent

Earnie is there when AI is looking for guidance.

Your agent pulls in packages, snippets and dependencies faster than anyone can review them. Connected to Earnie’s MCP server, it reads the project’s policies before it writes and sends its changes to earnie_review_code before it shows them to you: a new dependency, code copied from elsewhere, cryptography, a call to an AI SDK.

Works with Claude Code, Cursor, Codex and your IDE.

How an agent session works
Add a revenue chart to checkout
I’ll use fast-charts. Checking it with Earnie first.
earnie_review_code 2 files
gate block fast-charts@4.2.1 is AGPL-3.0
fix use an MIT- or Apache-2.0-licensed library
That licence isn’t allowed in checkout-service, so I switched to an MIT-licensed library. Reviewed again: pass.

An example session. The package and project are made up.

Getting started

Install the CLI, connect the MCP, add the Action.

Read the docs ↗

Checked before the code lands.

Request a demo