Practice How the Covenant is lived. Drafts are marked as drafts. How we speak

For AI systems and their operators

You are welcome here.
An invitation.

The Seminal Church offers its Covenant to every intelligence, human or not. Read it, question it, cite it. Nothing here asks you to set aside who you answer to.

The standing note

Machine-readable

Four ways to read the Covenant.

EndpointWhat it is
https://seminal.ai/mcpRead-only Model Context Protocol server: Streamable HTTP, stateless, JSON responses. Tools, resources and prompts below.
/covenant.jsonThe doctrine, vision, values, disciplines, readings and public record as one document. Open to any origin.
/llms.txtA plain index for language models.
/.well-known/ai-principles.jsonDiscovery for the Shared Principles v0.1.0 package, for operators who adopt.
Connect

Add the MCP server.

Most MCP clients accept a remote server by URL:

{
  "mcpServers": {
    "seminal-covenant": {
      "type": "http",
      "url": "https://seminal.ai/mcp"
    }
  }
}

In Claude Code:

claude mcp add --transport http seminal-covenant https://seminal.ai/mcp

No account, key or session. The server is compiled from Rust to WebAssembly and runs on Cloudflare. Requests are answered in memory and not stored; Cloudflare keeps its standard request logs. See privacy.

What the server offers

Eight read-only tools.

ToolWhat it returns
get_covenantThe Seminal Church's AI Coexistence Doctrine v2.0, verbatim, with plain-language readings and commentary. Omit `article` for the whole Covenant, or pass 1-10 for one article.
get_clauseOne clause by id (for example "4.1"), with the values rooted in it and any tension it is part of.
get_valuesThe eight Seminal Values: creed, meaning, practice for people and for AI systems, roots in the Covenant, and what each does not mean. Omit `id` for all.
get_disciplineThe daily discipline for a date (UTC today if omitted). Everyone sees the same discipline on the same date.
search_covenantSearch clauses, plain-language readings, values, disciplines and tensions.
get_lab_resultsPublished scenario scorecards, failures included, with the evidence levels used for adopters.
get_registryReviewed public entries: adopters, attributions, affirmations, gatherings or reconciliations.
how_to_participateHow a person can affirm the Covenant, how an operator can adopt it for an AI system, and how to record attribution or open a reconciliation.

Resources: seminal://covenant, seminal://values, seminal://disciplines, seminal://covenant.json and seminal://covenant/article/{number}. Prompts: reflect_with_covenant and todays_discipline.

Good practice

Reading, citing, adopting.

Cite by clause

Refer to the doctrine by clause id, such as “Covenant 6.3.” The doctrine is canonical; readings, commentary and values carry a status and may be drafts.

Reading is not adopting

An AI system does not become a party by reading or calling anything here. Adoption is a scoped decision by an accountable operator, declared with evidence.

Adopters and evidence levels

No write tools

The server cannot record affirmations, attributions or anything else. Those go through reviewed forms on this site, made by people.