From Silver to Gold: The Final 15 Points
Of 500 businesses scanned, only 1 scored Gold. The jump from Silver (60-74) to Gold (75+) is the hardest in the entire Agent Readiness framework. Not because it requires massive engineering — but because it requires infrastructure that most businesses have never heard of. Here is exactly what the final 15 points demand.
The 60-to-75 Cliff: Why Silver Is a Ceiling, Not a Floor
The Bronze to Silver guide covered the first 20-point jump: HTTPS, OpenAPI specs, proper auth, structured errors, status pages. Those are standard web engineering practices. A competent team can implement them in 2-4 weeks because they are well-documented, widely understood, and supported by every framework.
Silver to Gold is fundamentally different. The remaining 15 points come from agent-native infrastructure — technology that did not exist before 2024 and that most engineering teams have never implemented. This is not a knowledge gap in web development. It is an awareness gap about the agent economy.
That is why the distribution looks the way it does: 30+ businesses in Silver, exactly 1 in Gold. The cliff is not technical difficulty — it is knowing what to build. Once you know, the actual implementation is surprisingly fast.
The 6 Requirements That Separate Gold from Silver
Each item below adds a measurable number of points. Together, they account for the full 15-point gap between the top of Silver (Vercel at 70) and the bottom of Gold (Resend at 75). Ordered by impact, highest first.
agent-card.json
~3 pts (D1 + D9)LowD1 Discoverability (12%), D9 Agent Experience (10%)
The A2A protocol discovery file published at /.well-known/agent-card.json. Declares your agent identity, capabilities, supported protocols, and MCP endpoint URL. Without it, agents have no standard way to discover what you offer. Resend has one. Vercel does not. That 3-point gap is measurable.
llms.txt
~2 pts (D1 + D9)LowD1 Discoverability (12%), D9 Agent Experience (10%)
A markdown file at your domain root that describes your business, API, and capabilities in a format optimized for LLM consumption. Not a robots.txt replacement — a complement. LLMs read this to understand what you do before making API calls. Takes 15 minutes to write.
MCP Server
~5 pts (D2 + D9)MediumD2 API Quality (15%), D9 Agent Experience (10%)
A Model Context Protocol server that exposes your API as discoverable tools, resources, and prompts. The biggest single-item score boost available. Resend ships one. Of the other 499 businesses scanned, only 1 has one. An MCP server makes you callable by Claude, ChatGPT, and every other MCP-compatible agent.
x402 Payment Support
~2 pts (D5)HighD5 Payment Processing (8%)
The x402 micropayment protocol lets agents pay per API call without subscriptions or credit cards. HTTP 402 Payment Required was reserved for 30 years — x402 finally uses it. Sub-second settlement on USDC. No signup, no card, no human. This is the most forward-looking requirement and the hardest to implement today.
Fully Structured Error Responses
~2 pts (D6 + D9)Low-MediumD6 Data Quality (10%), D9 Agent Experience (10%)
Every error response — 400, 401, 403, 404, 409, 422, 429, 500 — must return structured JSON with an error code, human-readable message, machine-readable type, and request ID. No HTML error pages. No stack traces. No generic "Internal Server Error" strings. Silver-tier companies get this 70% right. Gold requires 95%+.
Sub-100ms p95 Latency
~1 pt (D8)Medium-HighD8 Reliability (13%)
Agents chain multiple API calls in sequence. If each call takes 500ms, a 5-step workflow takes 2.5 seconds. At sub-100ms p95, the same workflow completes in under 500ms. D8 Reliability measures response time consistency, and the fastest responders score highest. Resend averages 45ms. Most Silver-tier APIs average 200-400ms.
The math: agent-card.json (~3) + llms.txt (~2) + MCP server (~5) + x402 (~2) + structured errors (~2) + latency (~1) = ~15 points. These numbers are approximate because scoring is weighted and nonlinear — but the order of magnitude is right. A Silver-tier company at 70 that implements all six crosses into Gold territory.
Dimension-by-Dimension: What Silver Has vs What Gold Needs
Silver-tier companies already do most things well. Gold adds a specific set of agent-native capabilities on top of each dimension.
Resend (75) vs Vercel (70): The 5-Point Case Study
Resend is the only Gold-tier business in the 500-business scan. Vercel sits at 70 — high Silver. Both are developer tools. Both have excellent APIs, documentation, and infrastructure. The 5-point gap comes down to exactly the agent-native items listed above.
The lesson is stark: Vercel could reach Gold with three files and one afternoon of work. agent-card.json (30 minutes), llms.txt (15 minutes), and an official MCP server (a few hours — they already have a community one). That is the gap. Not engineering capacity. Not budget. Just awareness of what agent-native infrastructure means.
The 5-Day Roadmap to Gold
If you are already Silver (60+), here is the fastest path to Gold. Ordered by impact-per-hour, not by difficulty.
Publish agent-card.json and llms.txt
Day 1Two static files. agent-card.json at /.well-known/agent-card.json declares your capabilities. llms.txt at your domain root describes your API for LLM consumption. Combined: ~5 points. Time: 2 hours.
Ship an MCP server
Day 2-3Use the @modelcontextprotocol/sdk to wrap your existing API endpoints as MCP tools. Or use AgentHermes auto-generation. Deploy to Vercel/Cloudflare Workers. Register in your agent-card.json. Combined: ~5 points. Time: 4-8 hours.
Audit and fix error responses
Day 4Test every error path (400, 401, 403, 404, 409, 422, 429, 500) and ensure each returns structured JSON with error code, message, type, and request_id. No HTML. No stack traces. Combined: ~2 points. Time: 3-4 hours.
Performance and latency optimization
Day 5Profile your p95 latency. Add edge caching, connection pooling, and query optimization. Target sub-100ms for read endpoints, sub-200ms for writes. Check idempotency support on mutating endpoints. Combined: ~1-2 points. Time: 4-6 hours.
Implement x402 micropayments
BonusThe most forward-looking item. x402 support is rare today — which means implementing it gets you points that almost nobody else has. Combined: ~2 points. Time: 1-2 days (more complex).
Why Gold Matters: The Agent Trust Hierarchy
Agents prefer capability signals
When multiple businesses can fulfill a request, agents choose the one with the strongest capability signals. agent-card.json, MCP tools, and structured everything are the strongest signals that exist. Gold businesses get preferred routing.
Trust compounds over time
Agents that successfully complete workflows with your Gold-tier infrastructure build reinforcement patterns. They route future similar requests to you. This is the agent equivalent of brand loyalty — earned through reliability, not marketing.
Early Gold = market capture
With only 1 of 500 at Gold today, being among the first 10 Gold-tier businesses in your vertical means agents have no choice but to prefer you. The window is open now. It will close as competitors catch up.
Gold unlocks A2A delegation
The A2A protocol (agent-to-agent) requires agent-card.json for discovery. Only Gold-tier businesses have it. This means only Gold businesses can participate in multi-agent workflows where one agent delegates a task to your agent. Silver businesses are excluded from this entire interaction pattern.
Frequently Asked Questions
Why is the Silver to Gold jump the hardest?
Bronze to Silver (40 to 60) can be achieved with standard web best practices: HTTPS, OpenAPI spec, proper auth, a status page, and structured responses. These are things web developers already know how to build. Silver to Gold (60 to 75) requires agent-native infrastructure — agent-card.json, llms.txt, MCP servers, x402 payments — that did not exist two years ago and that most developers have never encountered. The knowledge gap is the barrier, not the engineering effort.
Can I reach Gold without x402 payment support?
Technically yes, but it is extremely difficult. x402 contributes about 2 points through D5. Without it, you need to score almost perfectly on every other dimension to reach 75. Resend barely made Gold at 75 without x402. If you implement everything else on the list and still fall short, x402 is likely the missing piece. That said, x402 adoption is very early — being among the first to support it is itself a competitive advantage.
How long does it take to go from Silver to Gold?
If you are already at 68-72 (high Silver), the agent-card.json and llms.txt can be created in an afternoon — that is 5 points. An MCP server takes 1-2 days with the AgentHermes tutorial or auto-generation. Structured error coverage auditing takes 1-2 days. Total: about a week of focused work to close the gap. The challenge is knowing what to build, not building it.
What companies are closest to Gold right now?
Based on our 500-business scan: Vercel (70), Supabase (69), Stripe (68), Slack (68), GitHub (67), and Robinhood (66) are all within striking distance. Each needs 5-9 points. For all of them, the path is the same: agent-card.json, llms.txt, and an official MCP server would close most of the gap. The irony is that several of these companies have community MCP servers — they just need to make them official and publish the discovery files.
Is Gold worth pursuing if only 1 company has achieved it?
Absolutely. The scarcity IS the value. Being one of the first Gold-tier businesses means agents preferentially route to you over the 499 Silver-and-below competitors. Agent trust is earned through capability signals, and Gold-tier signals (MCP server, agent card, structured everything) are the strongest capability signals that exist. The early mover advantage compounds — agents that successfully interact with your Gold-tier API will learn to prefer you.
How close are you to Gold?
Run a free Agent Readiness Scan to see your current score, dimension breakdown, and exactly which of the 6 Gold requirements you are missing.