API for developers

MCP server

An AI agent can use the same data through MCP at https://aiwriting.fyi/mcp. It takes POST only: a GET in a browser gives 405. The server uses Streamable HTTP, keeps no session and needs no key. It reads only. It has the same rate limit, the same public fields and the same credit line as the API. The server card describes it.

search_listings

: Search published AI writing tools by text, place and category.

get_listing

: Get one published AI writing tool by its slug.

list_places

: List countries, regions and cities that have AI writing tools.

list_categories

: List categories with their AI writing tool counts.

get_stats

: Get the number of published AI writing tools, places and categories, and when the data last changed.

list_sources

: List where the AI writing tool data comes from, with the license and refresh of each source.

list_courses

: List the published academy courses with their level, length, pass mark and certificate validity.

get_course

: Get one published academy course by its slug, with its objectives, lessons and FAQ. Exam questions and answers are never returned.

claude mcp add --transport http aiwriting https://aiwriting.fyi/mcp

{
  "mcpServers": {
    "aiwriting": {
      "url": "https://aiwriting.fyi/mcp"
    }
  }
}

curl -X POST "https://aiwriting.fyi/mcp" -H "content-type: application/json" -H "accept: application/json, text/event-stream" -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Responses

A successful response has data and meta. A list response also has page, with these fields:

limit

: items on each page.

total

: items that match.

nextCursor

: send it back as cursor, with the same filters and limit. It is null on the last page.

capped

: true when more items match than the cursor can reach.

fuzzy

: true when the items are close matches for q.

An error has error (a sentence), code and sometimes details. Branch on code: invalid_parameter, invalid_cursor, not_found, method_not_allowed, rate_limited, internal_error, unavailable.

Rate limits and caching

Each client may send 300 requests in 60 seconds. Every GET response has RateLimit-Limit and RateLimit-Policy. Over the limit you get 429 with Retry-After: wait that many seconds. The limit is counted at each Cloudflare location.

A successful response sends Cache-Control: public, max-age=60, s-maxage=300, stale-while-revalidate=600 and an ETag. Send the ETag back as If-None-Match to get 304 with no body.

Credit and licenses

Show this credit line next to the data you display:

Data from AI Writing (https://aiwriting.fyi). Each record keeps the license of its source, listed in meta.sources. A listing is a record, not an endorsement.

Every response lists its sources in meta.sources. Keep a source's license when you reuse its records.

• Listing enrichment: license source, refreshed manual
• Vendor websites, pricing pages and terms: license Each listing has a one-line summary of what the vendor says the product does, based on its own website, fetched 2026-11-13; some summaries echo the vendor's own wording. The home country comes from the vendor's own site (company address or governing-law clause) or, when the site does not state it, headquarters (P159) or country (P17) recorded on the company's Wikidata item; each listing's source note says which. Empty values mean the vendor did not state it., refreshed monthly
• Wikidata: license CC0, refreshed monthly