API.market Human-Dependent

Name: Agent Native Offer Audit — API.market
Item: API.market
Rating: 46
Author: Agent Native Offers

AUDIE Score: 46/100 · Audited 2026-04-11 · Website: https://api.market · Machine-readable: JSON

Pillar Scores

P1 Signal Architecture — 13/25

P2 Clarity Stack — 14/25

P3 Trust Envelope — 7/20

P4 Velocity Triggers — 6/10

P5 Gravity Design — 6/20

Executive Summary

API.market scores 46/100 (Human-Dependent), sitting in the middle of its tier — capable enough for developer discovery but not yet wired for autonomous agent evaluation and purchasing. Its strongest asset is the published MCP server, which represents a genuine step toward agent-native delivery. However, the absence of an llms.txt file means agents cannot systematically discover the 300+ API catalog, and the lack of scoped permissions, behavioral consistency guarantees, and a public status page means agents face unacceptable trust gaps when evaluating the platform as an autonomous buyer. The highest-ROI move is a /llms.txt file (low effort, +5 pts) followed by schema.org/Offer markup and a public status page to close the trust gap.

Strongest Signals

MCP Server Published An MCP server is live on Smithery and NPM, with a GitHub repo and setup documentation for Claude Desktop and Cursor. This is a meaningful step toward agent-native delivery — agents can discover and call APIs without manual setup.
Transparent Quantitative Rate Limits Per-API pricing includes specific calls/month and TPS figures, giving agents enough structured data to evaluate fit for a given use case before purchasing.
Instant Self-Serve Activation API keys are issued immediately after self-serve checkout with free tiers available — low friction for autonomous agents to get started.

Critical Gaps

No llms.txt The most glaring missing piece. With 300+ APIs, a well-structured llms.txt would allow AI agents to discover the full catalog and API capabilities without human intermediation. Currently returns 404.
No Behavioral Consistency Signals The terms page returns 404. No versioning policy, deprecation notice, or change-log found. An agent integrating a purchased API has no programmatic guarantee that terms or behavior won't change unexpectedly.
No Scoped Agent Permissions API keys are all-or-nothing. An agent buyer cannot acquire time-bounded, amount-capped, or action-specific permissions — a critical trust requirement for autonomous purchasing.
Zero Agent Memory or Compounding Value The platform has no memory of agent interactions, usage patterns, or personalized recommendations. Each session starts from zero, offering no gravity to retain agent buyers.

Priority Actions

Publish /llms.txt with full API catalog index — +5 pts · P1 · Effort: Low
Add schema.org/Offer markup to all API product pages — +4 pts · P1 · Effort: Medium
Create a public status page and link from homepage + API docs — +3 pts · P3 · Effort: Low
Publish a versioning and deprecation policy — +3 pts · P3 · Effort: Low
Introduce agent-scoped API key permissions — +5 pts · P3 · Effort: High

All 20 Criteria

P1-A Structured Data — 3/5
Homepage contains JSON-LD with `"@type": "Corporation"` (MagicAPI Inc), logo, and social links. Basic Organization schema only — no Product, Offer, or AggregateRating schema present.

P1-B Machine-Readable Pricing — 3/5
Individual API pricing is displayed in structured HTML tables with specific call limits and TPS rates (e.g., Basic: 5,000 calls/month, 2 TPS; Pro: 100,000 calls/month, 10 TPS). Not tagged with schema.org/Offer, but human- and moderately machine-readable structure exists.

P1-C llms.txt / Agent Layer — 0/5
/llms.txt returns 404. No agent identity layer or LLM-facing discovery file present.

P1-D API / MCP Availability — 4/5
Published MCP server on Smithery, NPM package (@noveum-ai/mcp-server), and GitHub repo (github.com/Noveum/api-market-mcp-server). Exposes 40 APIs from OpenAPI specs to LLMs. Documented setup for Claude Desktop and Cursor. REST API for marketplace interactions also available. Loses one point for the MCP server exposing only 40 of 300+ APIs and lacking a full agent card.

P1-E Discoverability (GEO) — 3/5
Listed on Smithery registry, ProductHunt, and third-party API directories. Blog content published. robots.txt has a sitemap reference. No specific GEO optimization for AI retrieval (no structured FAQ, no canonical LLM content layer).

P2-A Offer Completeness — 3/5
Product category (API marketplace), pricing, and API descriptions are findable. However, a complete machine-parseable picture of any individual API's offer requires navigating from the marketplace to the individual API page — not available from a single structured source.

P2-B Scope & Limits — 4/5
Rate limits are explicitly stated per API plan: calls/month, TPS (transactions per second), and hard/soft limit distinctions. Example seen: Trueway Routing API (Basic: 5,000/month, 2 TPS; Mega: 1,000,000/month, 30 TPS). Clear quantitative scope definition.

P2-C Substitution & Fallback Rules — 1/5
No guidance found on what happens when an API is unavailable, deprecated, or replaced. No fallback logic, substitution rules, or deprecation notices visible anywhere in audited pages.

P2-D Conditional Logic Transparency — 3/5
Most pricing is self-serve and visible without "contact sales" gates. Pricing tiers and access rules are relatively transparent. No major hidden conditions found on audited pages; terms page returned 404 preventing full assessment.

P2-E Semantic Precision — 3/5
Uses specific metrics: "99.9% uptime guarantee," "sub-2-hour response," "6M+ monthly requests," "10,000+ developers." Mix of precise stats and some vague marketing language ("trusted by developers worldwide").

P3-A Verifiable Performance — 2/5
Claims "99.9% uptime guarantee" and "SLA-backed enterprise infrastructure" with "real-time monitoring" stated on homepage. No public status page URL linked or visible. Self-reported only — no third-party verification (G2, Trustpilot, or uptime.is page found).

P3-B Scoped Permissions — 2/5
API keys issued per-API purchase with "encrypted API key management" mentioned. No evidence of time-bounded, action-scoped, or amount-limited permission tiers for agents. All-or-nothing access model per API.

P3-C Audit Trail — 2/5
"Detailed usage analytics" and "automated metering and billing" exist for human account holders. No agent-accessible transaction log API or machine-readable audit log found.

P3-D Behavioral Consistency — 1/5
No versioning policy, deprecation timeline, or notice period found. Terms of service page returned 404. No evidence of version-controlled API terms or stability track record communicated to buyers.

P4-A Friction-Free Activation — 4/5
"Instant API key activation" after self-serve checkout. Free tiers available (e.g., "5,000-1,000,000 free API calls" on select APIs). Live sandbox testing available before purchase. No human approval gate. Loses one point for requiring account creation and checkout flow.

P4-B Agent Decision Signals — 2/5
Free tiers and pricing visible. But no explicit machine-readable signals for when/why an agent should upgrade, subscribe, or act. No structured trial-expiry notification or programmatic trigger for autonomous decision-making.

P5-A Integration Depth / Switching Cost — 2/5
MCP server creates some integration stickiness, but exposes only 40 of 300+ APIs. Individual APIs are easily swappable for alternatives. No data lock-in mechanism or network effects creating significant switching cost.

P5-B Agent Memory / Personalization Layer — 1/5
No agent memory layer, usage history, or personalization features mentioned anywhere in audited pages. Each interaction appears stateless from the agent's perspective.

P5-C Programmatic Renewal Signals — 2/5
Automated metering and billing exist for human subscribers. No agent-accessible API for subscription renewal, usage threshold notifications, or auto-scaling signals.

P5-D Compounding Value Signal — 1/5
Marketplace catalog (300+ APIs) may grow over time, but no agent-readable signals indicating compounding value, usage-based improvements, or network effects visible to agents.

Rubric v1 (April 2026). Scores reflect the company's state on the audit date and may have improved since.