{ "company": "Requesty", "slug": "requesty", "website": "https://requesty.ai", "audit_date": "2026-04-16", "overall_score": 55, "tier": "Human-Dependent", "tier_as_published": "H", "pillars": { "P1": { "name": "Signal Architecture", "score": 12, "max": 25 }, "P2": { "name": "Clarity Stack", "score": 13, "max": 25 }, "P3": { "name": "Trust Envelope", "score": 11, "max": 20 }, "P4": { "name": "Velocity Triggers", "score": 9, "max": 10 }, "P5": { "name": "Gravity Design", "score": 10, "max": 20 } }, "criteria": [ { "id": "P1-A", "pillar": "P1", "name": "Structured Data", "score": 1, "max": 5, "evidence": "No schema.org markup (Organization, Product, Offer, AggregateRating) was detected on the homepage or pricing pages. The site has standard SEO metadata but no structured data that an AI agent could parse to evaluate the offer without rendering the full page." }, { "id": "P1-B", "pillar": "P1", "name": "Machine-Readable Pricing", "score": 2, "max": 5, "evidence": "Pricing is presented in a clear HTML feature table (Free/$0, Pro/5% markup, Enterprise/custom) with feature flags per tier. The 5% markup structure is human-readable prose, not a machine-tagged schema.org/Offer or JSON pricing feed. An agent can parse the table but cannot validate it as authoritative structured data." }, { "id": "P1-C", "pillar": "P1", "name": "llms.txt / Agent Layer", "score": 4, "max": 5, "evidence": "An llms.txt file is confirmed at docs.requesty.ai/llms.txt, with a referenced full version at docs.requesty.ai/llms-full.txt. Content describes Requesty as \"a unified LLM gateway and OpenAI-compatible API for 300+ AI models\" with endpoint details and integration guidance \u2014 directly structured for LLM consumption. Minor gap: llms.txt is on the docs subdomain, not the root domain (requesty.ai/llms.txt), which reduces discoverability." }, { "id": "P1-D", "pillar": "P1", "name": "API / MCP Availability", "score": 3, "max": 5, "evidence": "OpenAI-compatible REST API available at router.requesty.ai/v1 with documented endpoints (chat completions, messages, image generation, embeddings, model listing). Multiple SDK integrations confirmed (LangChain, Vercel AI SDK, PydanticAI, LlamaIndex). No published OpenAPI/Swagger spec found. No MCP server or agent card published. Functional but not fully agent-native." }, { "id": "P1-E", "pillar": "P1", "name": "Discoverability (GEO)", "score": 2, "max": 5, "evidence": "Developer-focused content (DataCamp tutorial, third-party blog coverage, AI SDK community provider listing). However, no evidence of LLM-optimized content strategy, structured entity pages, or deliberate generative engine presence beyond the llms.txt file." }, { "id": "P2-A", "pillar": "P2", "name": "Offer Completeness", "score": 3, "max": 5, "evidence": "The Free tier is fully defined ($0, $6 credits, 400+ models). The Pro tier pricing formula (5% markup) is stated but requires calculation to determine actual cost per provider. Enterprise is behind a \"Contact Sales\" gate with no published baseline. An agent can partially evaluate the offer but cannot determine full cost without model-provider look up." }, { "id": "P2-B", "pillar": "P2", "name": "Scope & Limits", "score": 2, "max": 5, "evidence": "No published rate limits, no documented token caps per key, no stated throughput maximums on public pages. Budget controls exist programmatically (user budget management, RBAC), but the parameters are not disclosed on public pricing or docs pages. Agents must make usage decisions without knowing the ceiling." }, { "id": "P2-C", "pillar": "P2", "name": "Substitution & Fallback Rules", "score": 2, "max": 5, "evidence": "Auto-failover is a core marketed feature (switches providers in <50ms), and the platform supports custom routing rules on Pro/Enterprise. However, the specific fallback chain \u2014 which providers substitute for which, under what conditions \u2014 is not documented publicly. An agent cannot pre-validate the substitution behavior before routing traffic." }, { "id": "P2-D", "pillar": "P2", "name": "Conditional Logic Transparency", "score": 2, "max": 5, "evidence": "Enterprise pricing conditions are opaque (contact sales, custom SLAs). Geo-routing behavior (EU \u2192 Frankfurt, US \u2192 Virginia, APAC \u2192 Singapore) is stated but the routing decision logic is not machine-readable. Conditions that trigger different pricing, routing, or governance behaviors are scattered." }, { "id": "P2-E", "pillar": "P2", "name": "Semantic Precision", "score": 4, "max": 5, "evidence": "Strong use of specific, verifiable metrics: \"8ms P50 latency overhead,\" \"99.99% uptime SLA,\" \"90 billion tokens processed daily,\" \"70,000+ developers,\" \"80% cost savings via caching,\" \"37% cache hit rate,\" \"12.3% average cost reduction.\" These are precise enough for an agent to use as evaluation criteria \u2014 well above industry average for LLM tools marketing." }, { "id": "P3-A", "pillar": "P3", "name": "Verifiable Performance Data", "score": 3, "max": 5, "evidence": "99.99% uptime SLA is published (self-reported). SOC2 compliance is stated. Third-party review presence confirmed on DataCamp, BestAITools, and Aitoolnet. Requesty is listed as a community provider in the Vercel AI SDK docs, which implies third-party validation. No dedicated public status page URL found \u2014 a gap for agent-runtime verification." }, { "id": "P3-B", "pillar": "P3", "name": "Scoped Permission Model", "score": 4, "max": 5, "evidence": "RBAC (role-based access control) with per-user budget limits, approved model whitelisting, and per-key spend controls are confirmed features. SSO/SAML on Enterprise. This is above-average agent-adjacent permission architecture, though no explicit \"agent scope\" (time-bounded, action-bounded token) is documented." }, { "id": "P3-C", "pillar": "P3", "name": "Audit Trail / Transaction Log", "score": 2, "max": 5, "evidence": "\"Audit logging capabilities\" are mentioned for Enterprise. Real-time analytics dashboard tracks usage by model, user, and team. However, no confirmation that audit logs are accessible via machine-readable API (versus human dashboard only). Enterprise-tier restriction limits baseline transparency." }, { "id": "P3-D", "pillar": "P3", "name": "Behavioral Consistency Signals", "score": 2, "max": 5, "evidence": "OpenAI-compatible API provides a stable interface anchor (changes to the upstream spec would require Requesty to adapt). However, no version-controlled terms of service, no published changelog, and no stated notice period for pricing or model changes found. The platform's youth (no long-term stability history) limits this score." }, { "id": "P4-A", "pillar": "P4", "name": "Friction-Free Activation", "score": 5, "max": 5, "evidence": "Free tier signup at app.requesty.ai with no credit card required. $6 in free credits issued immediately. API key available instantly upon registration. Integration is a single base URL change in existing OpenAI SDK code. This is textbook frictionless activation \u2014 an agent system could be configured to use Requesty with zero human gatekeeping." }, { "id": "P4-B", "pillar": "P4", "name": "Agent Decision Signals", "score": 4, "max": 5, "evidence": "Multiple programmatic decision signals present: free tier entry (zero-cost trial), explicit 5% markup pricing (cost predictable), published cache savings metrics (80% potential reduction), and model listing endpoint (agents can query available models). Minor gap: no explicit agent-legible \"tier upgrade trigger\" (e.g., \"upgrade when usage exceeds X tokens/day\")." }, { "id": "P5-A", "pillar": "P5", "name": "Integration Depth / Switching Cost", "score": 2, "max": 5, "evidence": "The platform is intentionally designed as a drop-in replacement with minimal lock-in (change one URL, change one key). While this is great for adoption, it means switching cost is structurally low. Integrations with LangChain/Vercel AI SDK/PydanticAI add minor stickiness via framework-level configuration." }, { "id": "P5-B", "pillar": "P5", "name": "Agent Memory / Personalization Layer", "score": 2, "max": 5, "evidence": "Usage analytics dashboards track historical spending, model performance, and cache rates by user/team. However, this data is dashboard-accessible for humans, not API-queryable for agents. No memory layer that an agent system can read to optimize its own behavior over time." }, { "id": "P5-C", "pillar": "P5", "name": "Programmatic Renewal Signals", "score": 3, "max": 5, "evidence": "API key management is fully programmatic via REST API (key creation, group management, organization admin). Budget controls can be set and adjusted via API. Pro-tier auto-renewal happens without human interaction once payment method is on file. Reasonably strong for renewal automation, though no explicit webhook or signal for \"renewal approaching\" events." }, { "id": "P5-D", "pillar": "P5", "name": "Compounding Value Signal", "score": 3, "max": 5, "evidence": "Cache hit rates compound as the same semantic queries are routed over time (37% average cache hit). Usage analytics become more predictive with volume. The routing intelligence improves as the platform learns provider performance. These are genuine compounding effects, but the signals are not surfaced in an agent-readable format that would justify autonomous re-commitment decisions." } ], "strongest_signals": [ { "title": "Near-perfect Velocity Triggers (9/10)", "detail": ": Requesty's free tier with instant API key issuance, no credit card required, and single-line integration is the clearest example of agent-native activation in this audit series. An autonomous agent system can begin using Requesty with zero human involvement." }, { "title": "llms.txt exists", "detail": ": Confirmed at docs.requesty.ai/llms.txt with a full version at docs.requesty.ai/llms-full.txt. This is a meaningful signal that the company understands LLM-native documentation, even if placement on a subdomain reduces discoverability." }, { "title": "Semantically precise metrics", "detail": ": Claims are stated in specific, verifiable terms (8ms P50 latency, 99.99% uptime, 80% cache savings, 90B daily tokens) \u2014 rare in the LLM tooling category and directly usable by agents as evaluation criteria." }, { "title": "RBAC + scoped access controls", "detail": ": Per-user budget limits, model whitelisting, and role-based access are better-than-average foundations for agent-scoped permissions, even without explicit agent scope tokens." } ], "critical_gaps": [ { "title": "No schema.org structured data", "detail": ": Homepage and pricing pages have no machine-readable Offer, Product, or AggregateRating markup. Agents that rely on structured signals for offer evaluation must fall back to HTML parsing \u2014 lossy and unreliable." }, { "title": "No published rate limits or caps", "detail": ": Without documented throughput limits, an autonomous agent system routing production traffic through Requesty cannot perform capacity planning or set safe rate controls. This is a significant gap for enterprise agent deployments." }, { "title": "No public status page", "detail": ": 99.99% uptime is claimed but unverifiable without a public status history. Agents operating in production need runtime-accessible uptime verification, not just marketing claims." }, { "title": "Low switching costs by design", "detail": ": The platform's biggest adoption advantage (drop-in replacement) is its biggest retention vulnerability. Any competitor offering a marginally better deal faces near-zero migration friction. For agents optimizing cost over time, this makes Requesty a permanently contested position." } ], "priority_actions": [ { "action": "Move llms.txt to root domain + expand it", "points_gain": 1, "pillar": "P1", "effort": "Low" }, { "action": "Add schema.org/Offer markup to pricing page", "points_gain": 3, "pillar": "P1", "effort": "Low" }, { "action": "Publish public status page + real-time uptime history", "points_gain": 2, "pillar": "P3", "effort": "Low" }, { "action": "Publish rate limits and throughput caps", "points_gain": 3, "pillar": "P2", "effort": "Low" } ], "executive_summary": "Requesty scores 55/100 in the Human-Dependent tier \u2014 the highest-scoring company in recent audit cycles when it comes to Velocity Triggers (9/10), where its instant-activation model is genuinely best-in-class. The platform's LLM gateway infrastructure is fundamentally aligned with agent-native demand: the thing agents need most (LLM API access with intelligent routing) is exactly what Requesty sells, and they sell it with near-zero activation friction. The critical gaps are all in the signal layer: no schema.org markup, no published rate limits, and a status page that doesn't exist publicly mean that agents evaluating Requesty as a production dependency cannot verify its claims without human research. The highest-ROI fix is adding schema.org/Offer markup to the pricing page \u2014 a 2-hour task that would move Requesty meaningfully toward the 65+ Emerging tier and signal to the agent-native buyer community that the infrastructure they run on is infrastructure that speaks their language.", "rubric_version": "v1-2026-04 (20 criteria, 100 raw points; P3-E Agent Registration added to rubric v2 in 2026-06, not scored in this audit)", "framework": "Agent Native Offers \u2014 The Agent Sale framework", "source_file": "2026-04-16 \u2014 Requesty \u2014 Agent Native Offer Audit.md", "rank": 18 }