Connect the MCP
One command adds the OES server to your coding assistant. It runs on your machine and stays up to date.
An open-source MCP for your AIBuild software
Open in Cursor
Build software
that respects its users.
Hold your project to the Open Ethics Standard, an open catalog of 800+ practices that mislead or harm users. Your AI finds the evidence. You decide what applies.
Add the MCP to your AI
claude mcp add --scope user open-ethics-standard -- npx -y open-ethics-standard-mcp@latestPaste into your terminal.
Free and open source · Runs on your machine · Needs Node.js 20+ · No account or API key
From code to considered review.
Your AI does the reading. The MCP supplies the standard. You make the call.
Your AI diagnoses
It reads your code, searches the catalog, and cites the file, lines, and evidence for every match.
You decide
Open the offline report, mark Applies, Unsure, or Dismiss, and export Markdown or PDF. Recheck after fixes.
Example: ask your AI for a review, and this is what it looks like.
Review our billing and signup flows against the Open Ethics Standard
⏺ open-ethics-standard · get_oes_catalog_overview 15 categories · 804 practices
⏺ Read src/billing/cancel.ts, web/consent/ConsentDialog.tsx, web/pricing/Offer.tsx
⏺ open-ethics-standard · search_oes_catalog("cancellation") → 2.1.3 Retention Dark Patterns
⏺ open-ethics-standard · create_oes_review(3 findings) → saved on this machine
3 possible matches in billing and signup.
Report: file:///…/reports/review.html
Possible matches3 findings · 2 files
2.1.3Retention dark patternsReview
src/billing/cancel.ts:84 — cancellation waits on a support chat.
Does this apply?AppliesUnsureDismiss
3.4.4Cookie-consent dark patternsReview
web/consent/ConsentDialog.tsx:31 — accept-all is prominent; no equivalent reject action found.
3.2.1Fake countdown timerReview
web/pricing/Offer.tsx:52 — countdown resets on every page load.
Then open your project and ask your AI: “Review my project against the Open Ethics Standard.” It replies with one link to the report.
The MCP runs on your machine, so your code and findings never leave it. Agent guide ↗
What your AI gets.
Your AI reads the code. The MCP gives it the standard and a review workflow, then turns its findings into your report.
get_oes_review_instructions
The review workflow, so every audit follows the same evidence rules.
get_oes_catalog_overview
The eight principles, 15 categories, and 804 practices, to plan coverage.
search_oes_catalog
Finds practices that match what the code actually does.
get_oes_practice
Full definitions, to verify each candidate before reporting it.
create_oes_review
Saves the offline report on your machine, so findings never leave it.
get_oes_practices
Batch definitions for every candidate in one call.
Backed by an open standard.
The Open Ethics Standard is a shared vocabulary for naming harmful practices. It isn’t a law, certification, or ethics score, and anyone can propose changes.
One source of truth
The MCP catalog is generated from a single Markdown standard.
Evidence with citations
Every finding points to a file, a line range, and an OES section.
Human judgment
AI can miss context, so you decide what applies.
Help shape the standard.
Suggest a practice, a correction, or a removal on GitHub.