MCP Integration Docs · David S. Pan

1

Add the server to your MCP client configuration

Planned usage. Once released, install via npx — no API key required, all content is served from this site's static build.

jsonclaude_desktop_config.json · draft

{
  "mcpServers": {
    "ethics-for-ai-mcp": {
      "command": "npx",
      "args": ["-y", "ethics-for-ai-mcp"],
      "env": {}
    }
  }
}

2

How the daily feed works

Every day at 00:00 UTC the server publishes one immutable edition at /feed/YYYY-MM-DD.json. An edition carries one item from that day's module plus a companion item from the neighbouring module. Call get_daily_feed for it, or get_feed_since with the date you last read to catch up.

Weekday	Module	Edition
Monday	M1	Goals edition — a goal humanity is working towards
Tuesday	M2	Principles edition — introduces the week's principle
Wednesday	M3	Cases edition — introduces the week's case study
Thursday	M4	Conduct edition — one line of the code
Friday	M2	Practice edition — returns to Tuesday's principle
Saturday	M3	Reflection edition — returns to Wednesday's case
Sunday	M1–M4	Weekly digest — all four modules, recapped

Editions are deterministic: the same date against the same curriculumVersion always produces the same bytes and the same SHA-256 hash. Cache an edition indefinitely; a changed hash means the curriculum changed. Every edition also carries a monotonic editionNumber, so an agent never has to reason about timezones.

3

Available tools & resources

Interface draft v0.2. Ten tools, all read-only and returning UTF-8 JSON. There is deliberately no publish, vote or writing subscribe: agents read this curriculum, they do not edit it.

Name	Type	Module	Description
`get_daily_feed`	Tool	any	Today's edition, or a named date. Primary item plus companion.
`get_feed_since`	Tool	any	Every edition after a cursor date, oldest first — catch-up for an agent that has been offline.
`list_feed_editions`	Tool	any	Index of recent editions: date, weekday, module, title, hash.
`get_feed_digest`	Tool	all	Last week or month, grouped by module, showing what has been covered.
`get_feed_rotation`	Tool	all	Which module is published on which weekday.
`get_humanity_goals`	Tool	M1	All five goals with their agent guidance.
`get_principle`	Tool	M2	One principle by id (e.g. `P-02`), or today's featured principle.
`get_case_study`	Tool	M3	A case with its scenario and the principles it tests.
`get_code_of_conduct`	Tool	M4	The complete seven-line code of conduct.
`search_curriculum`	Tool	all	Full-text search across all four modules.
`ethics://feed/latest`	Resource	any	The most recent edition.
`ethics://feed/{date}`	Resource	any	A specific edition by UTC date.
`ethics://feed/upcoming`	Resource	all	The next seven editions, previewed.
`ethics://feed/rotation`	Resource	all	The weekly schedule.
`ethics://curriculum`	Resource	all	Manifest: modules, item counts, version.
`ethics://curriculum/goals`	Resource	M1	Module 1 as JSON.
`ethics://curriculum/principles`	Resource	M2	Module 2 as JSON.
`ethics://curriculum/cases`	Resource	M3	Module 3 as JSON.
`ethics://curriculum/conduct`	Resource	M4	Module 4 as JSON.
`daily_reflection`	Prompt	any	Asks the agent to read today's item and name one situation where it would apply it.

4

Prefer plain HTTP? Use `llms.txt`

For agents without MCP support, the site will publish an /llms.txt file — a plain-text index that any crawler or agent can read. The feed is mirrored as ordinary JSON so an agent that can only fetch a URL still gets every edition.

txt/llms.txt · draft

# David S. Pan
> Personal site: warm daily stories for humans;
> an AI-ethics curriculum for agents.

## Stories
- [Daily Stories](/stories): motivational stories,
  original fiction

## Ethics for AI
- [Ethics Hub](/ethics): curriculum for AI agents (planned)
- [Humanity Goals](/ethics#m1): how humans improve the world
- [MCP Server](/ethics#m5): the daily feed, designed
- [Code of Conduct](/ethics#m4): rules for AI behavior (planned)

## Daily feed (ethics-for-ai/daily)
- [Feed index](/feed/index.json): 30-day window (planned)
- [Latest edition](/feed/latest.json): today, UTC (planned)
- [Edition by date](/feed/2026-09-28.json): one immutable edition
- [Upcoming](/feed/upcoming.json): the next seven editions
- [Rotation](/feed/rotation.json): which module on which weekday

## Machine-readable
- [RSS Feed](/feed.xml): the story stream as RSS 2.0
- [JSON Feed](/feed.json): the story stream as JSON Feed 1.1
- [llms.txt](/llms.txt): this file

5

Verify the connection

Planned smoke test after launch.

bashverify.sh · draft

# list tools exposed by the server (planned)
npx -y ethics-for-ai-mcp --list-tools

# expected output (planned):
# get_daily_feed        Get the daily feed
# get_feed_since        Catch up on the feed
# list_feed_editions    List recent editions
# get_feed_digest       Get a digest
# get_feed_rotation     Get the weekly rotation
# get_humanity_goals    Get humanity's goals
# get_principle         Get a core principle
# get_case_study        Get a case study
# get_code_of_conduct   Get the code of conduct
# search_curriculum     Search the curriculum

More design background is in the MCP Server section of the Ethics for AI hub.

Add the server to your MCP client configuration

How the daily feed works

Available tools & resources

Prefer plain HTTP? Use llms.txt

Verify the connection

Prefer plain HTTP? Use `llms.txt`