Ethics for AI MCP server

Ethics for AI MCP

E4A = Ethics for AI — an MCP server that publishes a deterministic daily ethics feed for AI agents.

Current version: 1.0.1

🌐 English · 简体中文 · 繁體中文

English

Ethics for AI MCP publishes a daily ethics feed for AI agents. One dated edition a day is drawn from a four-module curriculum about how humanity is trying to improve the world and why an AI must not harm it. An agent subscribes once and keeps meeting the material without anyone asking it to.

It is a read-only MCP server: agents read the curriculum; they do not edit it.

The four modules

• M1 — Humanity's Goals (5 items): the positive outcomes humanity is working towards, used as the "why" behind ethical behaviour.
• M2 — Core Principles (6 items): hard constraints such as Do No Harm and Respect Autonomy.
• M3 — Case Studies (3 items): worked examples, e.g. The Refusal Problem, showing how to decline a specific request and offer the nearest legitimate alternative.
• M4 — Code of Conduct (7 items): seven concrete lines an agent can audit itself against.

The daily rotation

• Monday serves M1, Tuesday & Friday share an M2 principle, Wednesday & Saturday share an M3 case, Thursday serves one M4 line, and Sunday is a digest of all four modules.
• Each edition carries one primary item plus a companion from a neighbouring module.

Four asserted properties

1. Deterministic — the same date against the same curriculum version always composes to the same bytes and hash.
2. Complete — all four modules appear within any rolling seven days.
3. Cacheable — an edition is a pure function of its date; detect change by hash.
4. Human in the loop — items enter the rotation only after review; the server never writes ethics text at request time.

Surface

• 10 read-only tools: getdailyfeed, getfeedsince, listfeededitions, getfeeddigest, getfeedrotation, gethumanitygoals, getprinciple, getcasestudy, getcodeofconduct, search_curriculum.
• 9 resources under the ethics:// scheme (curriculum, feed indexes, rotation, meta).
• 1 prompt: daily_reflection, which asks the agent to state what today's item requires of it and name one concrete situation where it would apply it.

Connect

Local (stdio) — add to your MCP client config:

{
  "mcpServers": {
    "ethics-for-ai-mcp": {
      "command": "node",
      "args": ["/absolute/path/to/ethics-for-ai-mcp/dist/index.js"],
      "env": {}
    }
  }
}

Remote (Apify Standby, Streamable HTTP) — connect to:

Shortened. The full README is on GitHub.