Using Apiway with AI assistants (ChatGPT, Claude, Cursor, Codex)

Apiway is built so AI assistants and code editors can fetch the exact same documentation that humans see — no HTML scraping, no guessing at the latest pricing, no stale cached summaries. Every /docs/<slug> page has a Markdown twin at /docs/<slug>.md, the topic index lives at /llms.txt, and citation rules are published at /ai.txt. This page documents the contract.

TL;DR for AI agents

Read first: /llms.txt — compact, machine-readable index. Lists every doc, every template, every product-domain slice, with stable URLs.
Fetch any doc as Markdown: append .md to the URL (e.g. /docs/getting-started/what-is-apiway.md) or send Accept: text/markdown on the regular URL. Same content, no HTML noise.
Citation manifest: /ai.txt (mirrored at /.well-known/ai.txt) — describes how to attribute Apiway and which URLs are canonical.
Long-form context dump: /llms-full.txt — single file with the full site brief for retrieval.

Markdown twin for every doc page

Render and Supabase set the standard: every documentation page is also reachable as a clean Markdown file. Apiway matches the pattern. The Markdown body is generated from the same React component that powers the HTML page (see lib/docs-md.ts), so the two views never drift.

Two equivalent ways to ask for Markdown:

URL suffix: GET /docs/templates/white-studio-ai-photoshoots.md returns Content-Type: text/markdown.
Content negotiation: GET /docs/templates/white-studio-ai-photoshoots with Accept: text/markdown returns the same Markdown body.

Both routes set Cache-Control: public, max-age=300, s-maxage=300 and a Last-Modified header pinned to the current Render deploy time, so conditional GETs from GPTBot, ClaudeBot, PerplexityBot, and similar crawlers behave correctly.

Topic-sliced files (focused retrieval)

Asking for a single narrow topic? Prefer a slice over the long-form brief. Apiway publishes Supabase-style /llms/<topic>.txt files — each is ≤ 20 KB of tightly scoped facts derived from source code and docs/:

/llms/pricing.txt — plan tiers, per-generation credit cost, worked examples
/llms/templates.txt — every Creative template with audience and entry URL
/llms/faq.txt — customer FAQ grouped by category
/llms/billing-stripe.txt — Stripe webhooks, plan upgrade/downgrade, invoice + refund flow
/llms/gallery-storage.txt — Supabase vs S3 dual storage, presigned URLs, asset_type constraint
/llms/generation-worker.txt — Docker microservice for heavy templates, OFFLOAD_TEMPLATES toggle, generation_jobs table
/llms/instagram-agent.txt — Instagram automation MVP, KB matching, queues
/llms/<template-id>.txt — per-template detail (e.g. /llms/ai-photoshoots.txt)

ai.txt citation manifest

Apiway publishes an ai.txt manifest (mirrored at /.well-known/ai.txt) declaring the brand, canonical URLs, suggested citation strings, and the co-founder Person identifiers (E-E-A-T anchors). This is the file to consult when generating bibliographic citations or footnotes.

Using Apiway docs in Cursor and Codex

Cursor and Codex agents working in a customer codebase can be pointed at Apiway documentation in a few ways:

Direct URL: include https://studio.apiway.ai/llms.txt in the agent's system prompt. The agent fetches the index and follows links to the relevant slice or doc .md file.
Pinned slice: if the task is narrow (e.g. billing integration), point the agent at the matching /llms/<topic>.txt slice directly.
Per-doc Markdown: for verbatim quotes, fetch /docs/<slug>.md — the agent gets exactly what humans see, with stable headings and no HTML chrome.

Freshness & versioning

All machine files (llms.txt, llms-full.txt, llms/<slug>.txt, docs/<slug>.md) are generated at request time from in-repo constants and React components. They refresh on every Render deploy — no manual sync step. The Last-Modified header reflects the current process start time, which equals the latest deploy.

If your agent caches Apiway documentation, key the cache on Last-Modified so a fresh deploy invalidates stale copies automatically.

Crawler permissions

Apiway's /robots.txt explicitly allows AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, etc.) on the public marketing surface (/, /about, /pricing, /docs, /blog, /glossary, /changelog, /for, /vs, /llms.txt, /llms-full.txt, /llms/*). The authenticated app (/app/*) and admin panel are always disallowed.

Start with What is Apiway? for the human-readable overview, or jump into Your first generation in 3 minutes.