# ING — API Agent Score

> Score: 62/100 (Grade: C) | Domain: ing.com | Rubric: 1.0.0 | Checked: August 19, 2026

ING scored 62/100 (C), classified "partially-ready".

[View full report](https://www.postman.com/ai/ai-ready-apis/company/ing)
[All organizations](https://www.postman.com/ai/ai-ready-apis/llms.txt)

## Summary

- Overall: 62/100 (C)
- Classification: partially-ready

## Category Scores

- API Design: 88/100 (A)
- Developer Experience: 49/100 (F)
- Agent Discovery: 61/100 (C)
- Agent Understanding: 76/100 (B)
- Agent Usability: 45/100 (F)

## Check Results

### API Design

- [warn] Auth declared & discoverable — Authentication is declared on only 6 of 20 operations, so an agent can't tell which endpoints need credentials. Investigated: spec 90%, docs 80%, wellknown 0%.
- [warn] Schema coverage & depth — 3 of 125 public signatures use loose types (anyTypeRatio=0.024). Investigated: sdk 90%, spec 67%.
- [warn] Machine-readable, versioned contract — The API sets a version ("5.0.1") but exposes no versioning scheme in the URL, a header, or the media type, so an agent can't pin to a specific version. Investigated: spec 85%, docs 70%.
- [pass] Security & governance hygiene — No credential-shaped strings detected in spec. Investigated: spec 100%, sdk 100%, wellknown 0%.
- [warn] Example coverage — README is present but has no runnable quickstart. Investigated: sdk 50%, spec 0%, docs 0%.

### Developer Experience

- [pass] Description completeness — Only 83% of operations and parameters have substantive descriptions (6 of 9 ops, 48 of 48 params; target 90%+). Investigated: spec 100%, sdk 76%.
- [warn] Quickstart present — Quickstart page is reachable at https://developer.ing.com/openbanking/quickstart (12 variants scanned) but has no runnable code sample in its HTML or .md variant, so there's nothing to copy and run. Investigated: docs 50%.
- [fail] Self-service developer portal — No self-service signup detected — no signup link on the homepage or docs, and no conventional signup path (/signup, /sign-up, /register, /get-started, /console/signup, /dashboard/signup, /try, /try-free, /free, /free-trial, /start, /start-free, /join, /create-account, /account/signup, /auth/signup, /users/sign_up) resolved — so an agent can't onboard on its own. Investigated: docs 0%.
- [fail] Code samples in docs — No code samples detected across 1 sampled docs pages, so a coding agent gets no ready-to-use examples. Investigated: docs 0%.
- [fail] Changelog published — No changelog link or mention appears in the spec, so an agent can't tell what changed before something breaks. Investigated: spec 0%, docs 0%, sdk 0%.

### Agent Discovery

- [pass] Registry & SDK presence — Indexed on Context7 (websites/devportal_ing_pl, 5 snippets). Investigated: docs 100%, sdk 67%, cli 0%, mcp 0%, wellknown 0%.
- [warn] Docs reachable, not hard auth-gated — 1 JavaScript redirect(s) detected across 1 pages. Investigated: docs 50%.
- [warn] Crawlable / AEO — No sitemap found via robots.txt or /sitemap.xml on the docs host, so AI search engines have no crawl map for your docs. Investigated: wellknown 50%.
- [fail] llms.txt present, valid & comprehensive — No llms.txt found at any candidate location (https://developer.ing.com/openbanking/llms.txt, https://developer.ing.com/llms.txt, https://developer.ing.com/docs/llms.txt). Investigated: docs 0%, sdk 0%.

### Agent Understanding

- [pass] Machine-readable errors (RFC 9457) — Only 2 distinct 4xx/5xx error codes are documented (plus 0 catch-all "default" responses), so an agent has limited failure branching. Investigated: docs 100%, sdk 100%, spec 50%.
- [warn] Operation purpose clarity — Read/write classification is partial: 88/133 classified (46 read, 87 write). Investigated: sdk 83%, spec 40%.
- [warn] Agent-navigable, token-efficient docs — No pages support .md URLs (0/1 tested). Investigated: docs 79%.
- [warn] Agent instructions file (AGENTS.md) — No agents.md or skill.md context file found (nice-to-have for agent operation). Investigated: sdk 50%, wellknown 0%.
- [warn] Description consistency across surfaces — Mean pairwise description similarity across 3 surfaces (spec, docs, sdk) is 1% (threshold 35% for full credit).
- [fail] Docs structured data — No JSON-LD or OpenGraph/meta tags found across 1 assessed pages (1 JS-rendered), so answer engines have nothing to cite. Investigated: docs 0%.

### Agent Usability

- [warn] Sandbox separation — No test-mode flag or sandbox environment is exposed. Investigated: sdk 50%, spec 0%, docs 0%.
- [warn] Runnable collection with test scripts — README is present but has no runnable quickstart. Investigated: sdk 50%, platform 0%.
- [fail] Idempotency documented — Only 0% of mutating operations (0 of 9) document idempotency, so an agent's retries can create duplicate writes. Investigated: spec 0%, docs 0%, sdk 0%.
- [fail] Rate-limit signaling — No rate-limit response headers are documented, so an agent can't tell when it is approaching a limit and will get throttled. Investigated: spec 0%, docs 0%.
- [skip] Pagination documented & consistent — No surface produced evidence for this capability in this run.

## Executive Summary

ING's API Design is a genuine strength — the contract is well-shaped, typed, and secured, giving partners a solid foundation to build against. However, two areas are actively blocking partner program growth: operational reliability gaps (no idempotency or rate-limit guidance) mean AI agents will create duplicate writes and get throttled without warning, while the developer experience lacks self-service onboarding and code samples, forcing partners through manual processes before they can make a single call. Closing these gaps — alongside adding a changelog, llms.txt, and structured metadata — will unlock self-serve partner adoption and make ING's APIs safely consumable by both human developers and AI agents.