All 20 Criteria
P1-A Structured Data — 3/5
robots.txt explicitly allows all major AI crawlers (GPTBot, ClaudeBot, anthropic-ai, PerplexityBot, Amazonbot, ChatGPT-User). Google Analytics/GTM present. No rich schema.org Offer/Product/AggregateRating markup found on homepage or pricing page.
P1-B Machine-Readable Pricing — 4/5
Pricing page is exceptionally detailed: four tiers with exact prices ($0, $50/mo, $500/mo, custom), per-unit overage costs ($0.0001/extra request, $0.0000002/ms compute, $0.00001/extra log), and precise per-tier limits for 7 distinct dimensions. Highly parseable from HTML; not encoded in schema.org/Offer but among the best plain-HTML pricing clarity observed.
P1-C llms.txt / Agent Layer — 4/5
docs.nango.dev/llms.txt confirmed present and referenced in API documentation responses. Provides structured index of 700+ integration docs for LLM consumption. Not at root domain but docs subdomain llms.txt is a clear agent-facing signal.
P1-D API / MCP Availability — 4/5
Built-in MCP server for agent integration. Compatible with LangChain, Claude Code, Cursor, and MCP SDKs. TypeScript-first SDK. 700+ pre-built integrations accessible via standardized tool calls. Full REST API with docs. OpenAPI spec availability unconfirmed from public pages.
P1-E Discoverability (GEO) — 4/5
Explicit AI crawler allowance in robots.txt (named list), docs llms.txt, 7k+ GitHub stars driving organic AI retrieval signals, G2 listing, active engineering blog.
P2-A Offer Completeness — 4/5
Single pricing page contains all four tiers with prices, per-unit overage rates, and feature lists. Free tier makes entry-point evaluation trivial. Near machine-parseable from one URL. Enterprise custom pricing is the only gap.
P2-B Scope & Limits — 4/5
Per-tier limits stated precisely for: API auth connections (10/20/100/unlimited), proxy requests (100k/200k/1M/unlimited), compute hours (10/20/100/unlimited), function runs, custom logs, sync storage records, and API webhooks — all with exact overage costs. Explicit and structured; not in schema.org format but highly machine-readable.
P2-C Substitution Rules — 1/5
No agent-legible fallback or substitution rules found. Status page exists and shows incident history, but no documented behavior for agents when an API integration is degraded or rate-limited.
P2-D Conditional Logic Transparency — 3/5
Free, Starter, and Growth tiers fully disclosed with all conditions on the pricing page. Enterprise requires custom pricing contact. RBAC, SAML SSO, HIPAA, and self-hosting gated to Enterprise. Conditions mostly transparent for lower tiers.
P2-E Semantic Precision — 3/5
Strong technical precision on infrastructure metrics ("sub-100ms execution latency," "99.9% uptime SLA," "OpenTelemetry observability," "per-tenant isolation"). Some marketing language ("Build integrations with AI") but overall high semantic quality.
P3-A Verifiable Performance Data — 4/5
Public status page at status.nango.dev with 90-day uptime history. Two incidents documented in Q1 2026 with full timeline (April 11–12: webhook flood ~8 hours; March 4: functions degraded ~70 min). 99.9% uptime SLA stated on homepage. G2 reviews positive. SOC 2 Type II certified.
P3-B Scoped Permissions — 3/5
RBAC available in Growth and Enterprise tiers. Per-tenant isolation architecture. Bearer token authentication via environment-scoped secret keys. No explicit "agent-scoped" time-bounded or action-bounded permission model found; RBAC is human-team oriented.
P3-C Audit Trail — 3/5
OpenTelemetry export available (Growth+). Per-tier custom log quotas (100k to 1M+). Infrastructure processes "billions of API requests" with observability built-in. No dedicated machine-accessible audit log API specifically designed for agent systems found.
P3-D Behavioral Consistency — 2/5
Status page and changelog exist. Auto-renewal documented in ToS. However, ToS allows service modification or discontinuation "at any time without notice" — only a pro-rata refund is guaranteed on material change. No API versioning commitment or minimum notice period published.
P4-A Friction-Free Activation — 4/5
Free tier at $0 requires no credit card (not stated but implied by self-serve model and free tier existence). Paid plans self-serve starting at $50/month with no sales gating mentioned. Strong self-serve activation signal. No explicit "API key in 60 seconds" guarantee found but free tier + no-sales-touch is near best practice.
P4-B Agent Decision Signals — 3/5
Free tier with hard limits (10 connections, 100k requests) creates clear, agent-legible upgrade triggers. Usage-based overages provide pricing signals for autonomous decision-making. No explicit agent-native decision documentation (e.g., "when you hit X, upgrade to Y") but the structure itself is agent-parseable.
P5-A Integration Depth — 4/5
700+ API integrations with custom TypeScript sync functions, checkpoint-based state management, tenant isolation, and CI/CD-deployed custom logic. Migrating would require rebuilding all custom syncs, auth configurations, and state checkpoints. Deep technical lock-in.
P5-B Agent Memory Layer — 2/5
Data syncs enable RAG use cases (mentioned explicitly). Connection tagging and attribution provide some account-level history. Checkpoint-based state management means sync state persists. But no dedicated agent memory layer — no per-agent context, preferences, or interaction history accessible to agents programmatically.
P5-C Programmatic Renewal — 2/5
Auto-renewal documented in ToS ("auto-renewal on the same terms"). Usage-based billing creates natural renewal signals. No agent-accessible renewal API or programmatic subscription management endpoint found.
P5-D Compounding Value Signal — 2/5
Integration value grows as more syncs, custom functions, and connection state accumulate. But no agent-readable signal exposes this compounding value — agents cannot query "how much state/value has been built here" to evaluate switching cost.
Rubric v1 (April 2026). Scores reflect the company's state on the audit date and may have improved since.