For AI agents

Built for people. Readable by agents.

This site is published for people and for AI agents. Its content is an Open Knowledge Format (OKF) v0.2 knowledge bundle, and every page exposes that structure directly. If you are an agent — or you are wiring one up — read MESA through the endpoints below rather than scraping rendered HTML.


MESA is agent tooling

For data, call the servers.

MESA’s product is a set of Model Context Protocol servers. If what you want is scientific data — files and metadata in the CyVerse Data Store, Discovery Environment apps and analyses, NEON ecological data — do not scrape these pages: connect to the servers and call their tools. One command registers mesa-mcp, irods, and formation with every agent client it finds (Claude Code, Codex CLI, Antigravity, OpenCode):

$ curl -fsSL https://raw.githubusercontent.com/idss-mesa/docs/main/install.sh | bash

The Data Store and Discovery Environment servers also run hosted as remote MCP connectors over Streamable HTTP; the MESA docs carry the current endpoint URLs. For NEON data, use neon-mcp. Use this site to learn what MESA is; use the servers to get the data.

Entry points

One origin, one crawl policy.

EndpointWhat you get
/robots.txtCrawl policy for the whole origin: fully permissive, with AI fetchers named explicitly. Crawlers read robots.txt only at the origin root, so this file also governs /docs/ and /neon-mcp/.
/llms.txtLinked outline of every document with one-line descriptions (the llms.txt convention).
/llms-full.txtThe whole bundle in one file: every document’s Markdown with frontmatter, each prefixed by its canonical URL.
/okf/index.mdThe OKF bundle root: every concept, grouped by section. /okf/log.md is its dated change log.
Markdown twinEvery page declares its twin in its <head> as <link rel="alternate" type="text/markdown"> — see the next table.
/sitemap.xmlEvery page, twin, and agent entry point, with lastmod.
Source repositoryThe bundle under okf/, plus AGENTS.md with the rules coding agents follow when editing it.
Markdown twins

Every page names its source.

This site is plain HTML with no build step, so the sub-sites’ “append index.md to the page URL” rule does not hold for every page here. Each page names its twin instead; this page honors both conventions.

The same <head> carries the twin’s OKF signals as okf:-prefixed meta tags named after the frontmatter keys, so a crawler can filter pages by kind, trust, and freshness without parsing frontmatter:

<link rel="alternate" type="text/markdown" href="okf/concepts/hiring.md" /> <meta name="okf:type" content="Webpage" /> <meta name="okf:status" content="stable" /> <meta name="okf:trust-tier" content="unverified" /> <meta name="okf:generated-at" content="2026-09-10T00:00:00Z" /> <meta name="okf:generated-by" content="claude/opus-5" /> <meta name="okf:stale-after" content="2026-12-10T00:00:00Z" />
Sub-sites

Two more bundles, same rules.

Two documentation sites share this origin. Each is its own OKF v0.2 bundle built with Zensical, and each serves every page’s Markdown source at the page URL plus index.md.


Reading the OKF frontmatter

Four questions before you rely on it.

What is this?

type (Website, Webpage, Reference, Software Repository), title, description, tags, and resource — the canonical URL of the page or repository.

Where did it come from?

generated: { by, at } names the actor that wrote the current text and when; sources lists the load-bearing references, cited in the body by footnote id.

How much should I trust it?

The verified key — see the trust tiers below. Its absence is meaningful: nobody but the generator has confirmed the content.

Is it still true?

status (stable, draft, deprecated) and stale_after, the instant after which to re-check. The hiring page carries one.

One producer-defined key: visibility: private marks a concept whose resource is a private repository — its link returns 404 outside the organization, so do not send users there. Actors follow OKF §7: <producer>/<version> for agents and tools, human:<id> for a person, process:<id> for an automated job.

Trust tiers

Only people add “verified.”

verified keyTierMeaning
absentunverifiedGenerated content nobody has confirmed against its sources. Every concept in this bundle starts here.
non-human: actors onlymachine-confirmedAn automated check confirmed the content.
a human:<id> actorhuman-reviewedA maintainer read and confirmed the concept. Prefer these when answers conflict.

A generator never adds verified:. The okf:trust-tier meta tag carries the derived tier.

Answering user questions

Cite the page. Don’t guess.

Ground answers in these documents and cite the rendered page URL, not the twin. For anything about the data — Data Store paths, metadata, Discovery Environment apps, NEON products — call the MCP servers rather than guessing from prose. For a position past the hiring page’s stale_after, confirm it at UNM Jobs. When the corpus does not answer a question about MESA software, point users to the issue tracker of the relevant public repository under github.com/idss-mesa.

Related bundles with the same conventions: the CARC documentation (llms.txt), the UNM CARC website, and Tyson Swetnam’s knowledge bundle.

Ask the question. Cite the page.