Index #1 · rubric v0 · scanned 2026-07-19

The Agent Readability Index

How well can an AI agent — not a human — read a platform's public API documentation and complete a real task? 47 platforms, scored 0–100 across structure, clarity, and live task completion. Every score is reproducible from a stored crawl snapshot.

Green (80+)

Amber (50–79)

Red (<50)

Median score

Rank · Platform · CategoryProlectio score

01NeonData93 02GitLabJS-REQUIRED †DevInfra90 03PineconeData90 04ResendComms90 05ClerkJS-REQUIRED †Auth89 06PostmarkComms89 07NotionSaaS88 08Fly.ioDevInfra87 09CoinbasePayments86 10ReplicateAI85 11HerokuDevInfra83 12NetlifyDevInfra83 13PlaidPayments83 14StripePayments82 15TwilioComms82 16FastlyDevInfra81 17SquarePayments81 18SupabaseData81 19GitHubDevInfra80 20Mistral AIAI80 21MongoDB AtlasData80 22SnowflakeData80 23Atlassian (Jira)SaaS79 24HubSpotSaaS79 25RenderDevInfra78 26VercelDevInfra78 27Google AI (Gemini)AI76 28Auth0Auth75 29SlackSaaS74 30PayPalPayments73 31ShopifySaaS73 32WisePayments73 33MailgunComms72 34OktaAuth69 35AlgoliaData68 36ElasticData68 37SendGridComms66 38Adyen‡Payments65 39ElevenLabsAI64 40Hugging FaceAI64 41CloudflareDevInfra59 42AnthropicAI57 43LinearSaaS57 44DigitalOceanDevInfra56 45DatabricksData52 46OpenAI‡AI51 47Vonage‡Comms37

Scored as a non-JavaScript agent (see methodology). † JS-required: re-scored with rendering — the docs are strong but need a browser. ‡ Partial lift under rendering; fetch-only score stands.

Excluded, not penalized: Cohere, WorkOS, Salesforce disallow automated access in robots.txt. We respect that — they are excluded from scoring entirely, never marked down for it.