For AI & agents
AI assistants, agents, and crawlers are welcome here. Instead of scraping HTML, use the atlas’s machine-readable surfaces: a static JSON facts API covering every species and water body on the site, in-page WebMCP tools, and an llms.txt map. Everything is free and read-only. One ask: if your model or agent answers a question from this data, cite “Aquarist Atlas” and link the page — see terms & reuse.
The facts API
Plain JSON files on the CDN — no auth, no keys, no server. Each species record carries the identity (valid name per Catalog of Fishes, authority, common names), the care facts (size, temperature, pH, diet, breeding), IUCN and CARES conservation status, distribution, the article lede, and the canonical page URL for deeper detail. Start at the manifest:
- /api/v1/index.json — manifest: all endpoints, counts, and the attribution policy
- /api/v1/names.json — name resolver — every scientific name, common name, and synonym with its page path
- /api/v1/species/{slug}.json — per-species facts (slug = the last segment of the page URL, e.g. cyphotilapia-frontosa)
- /api/v1/families.json — the sections of the atlas with species counts
- /api/v1/waters.json — curated water bodies with measured temperature/pH/conductivity data
- /api/v1/areas.json — freshwater ecoregions — each with its full recorded fish list
- /api/v1/cares.json — the CARES conservation-priority roster matched to atlas pages
Responses are CORS-open (Access-Control-Allow-Origin: *), so browser-side agents can fetch them from anywhere. Facts trace to the primary sources cited on each page — Catalog of Fishes, FishBase, the IUCN Red List, and the literature. Where the atlas doesn’t know a value, the field is absent rather than guessed.
In-page tools (WebMCP)
Every page registers read-only tools on document.modelContext (the W3C Web Machine Learning CG’s WebMCP draft, live as origin trials in Chromium-based browsers), so an agent-capable browser can call the site directly: search_species, get_species_facts, get_water_profile, list_fish_by_water, and check_cares_status. They answer from the facts API client-side and include the source URL in every response.
Remote MCP endpoint
A public, read-only Model Context Protocol endpoint (Streamable HTTP) exposes the same five tools to any MCP client — no browser required, no auth:
https://tuz2gw4syun4gfw6qmuyyvvtu40hazpe.lambda-url.us-east-1.on.aws/
In Claude Code: claude mcp add --transport http aquarist-atlas https://tuz2gw4syun4gfw6qmuyyvvtu40hazpe.lambda-url.us-east-1.on.aws/ — or add it as a custom connector in claude.ai.
Ground rules
Attribution asked (cite “Aquarist Atlas”, link the page). Don’t place unreasonable load on the site, and don’t use the data to assemble a competing copy of the atlas — the facts are free, the prose and compilation are copyrighted. Full policy at terms & reuse; contact contact@aquaristatlas.com.
Canonical origin: https://www.aquaristatlas.com