Why Tally and Growthbook Both Score 64-65: The Developer Tool Silver Plateau
Tally builds forms. Growthbook manages feature flags. Different products, different markets, different codebases. Yet they score within one point of each other on agent readiness: Tally at 65, Growthbook at 64. This is not a coincidence. It is a pattern we see across every well-built developer tool that has not adopted agent-native features. We call it the Silver Plateau — the ceiling where good APIs get stuck without three specific files.
Score Comparison: Two Products, One Ceiling
The side-by-side scores tell the story. Both tools excel at the fundamentals — API quality, onboarding, documentation, pricing — and both collapse at D9 Agent Experience. The pattern is so consistent that you could predict the total score from D9 alone: if D9 is under 20, the tool scores Silver regardless of how good everything else is.
Tally
Growthbook
The pattern: D1 through D8 are strong for both. D9 Agent Experience is catastrophically low — 18 and 15 respectively. This single dimension drags the entire score from Gold territory (75+) down to Silver (60-74). The Silver Plateau is, fundamentally, a D9 problem.
What Both Tools Get Right
These are not bad products with low scores. These are excellent products that happen to be missing the agent-native layer. Understanding what they do well clarifies why the ceiling exists — and how close they are to breaking through.
Clean REST APIs
D2 API Quality: 75-78Both have well-designed REST APIs with consistent URL patterns, proper HTTP methods, and structured JSON responses. This is the foundation that gets them to Silver.
Self-Service Onboarding
D3 Onboarding: 80-82Sign up, get API keys, start building — no sales call required. Both offer free tiers with generous limits. Documentation is comprehensive and example-rich.
Good Documentation
D1 Discovery: 70-72Both maintain dedicated docs sites with guides, API reference, and code examples. Growthbook has the additional advantage of open-source documentation on GitHub.
Transparent Pricing
D4 Pricing: 68-70Clear pricing pages with feature comparison tables. Free tiers that let developers evaluate without commitment. No "contact sales" gates on standard plans.
This profile — strong API, strong docs, strong onboarding, weak agent-native protocols — is the defining characteristic of the Silver Plateau. It describes not just Tally and Growthbook but dozens of developer tools we have scanned. The tools are built for human developers, not AI agents. That was the right strategy in 2023. In 2026, it is leaving points on the table.
The Three Gaps That Define the Ceiling
The Silver Plateau is not a vague problem — it is three specific missing files. Every developer tool stuck at 60-65 is missing the same three things. Fix them and the score jumps 25-35 points.
No agent-card.json
+8-12 pointsNeither has a /.well-known/agent-card.json file. AI agents discovering tools have no structured way to understand what Tally or Growthbook offers or how to interact with their APIs.
Fix: Add a 20-line JSON file at /.well-known/agent-card.json describing capabilities, auth, and endpoints.
No MCP Server
+10-15 pointsNeither offers an MCP server. Agents must manually construct API calls from documentation rather than discovering and calling tools through a standard protocol.
Fix: Wrap existing API endpoints as MCP tools. Tally: create_form, get_submissions. Growthbook: get_features, create_experiment.
No llms.txt
+5-8 pointsNeither has a /llms.txt file explaining their capabilities in natural language. LLMs building integrations have to parse HTML docs rather than reading a purpose-built machine context file.
Fix: Add a text file at /llms.txt describing the product, API capabilities, and common use cases in plain English.
This is the insight that makes the Silver Plateau actionable: the gap is not about better documentation, faster APIs, or more features. It is about three files that did not exist when these products were designed. Adding them is a weekend project for a single developer. The Craft case study showed the same pattern — excellent product, missing agent layer.
The Three-Step Breakout: Silver to Gold in a Weekend
The Silver to Gold guide covers this in detail, but here is the specific playbook for developer tools like Tally and Growthbook. The hard work — building a quality API — is already done. Breaking through the plateau requires adding the agent-native discovery layer on top.
Add agent-card.json (30 minutes)
Create /.well-known/agent-card.json with product description, API endpoints, auth method, and capability list. This is the minimum viable agent discovery file.
+8-12 points → Score: 72-77
Add llms.txt (1 hour)
Create /llms.txt explaining your product, API capabilities, authentication, and common integration patterns in plain English. LLMs read this when building integrations.
+5-8 points → Score: 77-85
Ship an MCP Server (1-2 days)
Wrap your top 5 API endpoints as MCP tools. For Tally: create_form, list_forms, get_submissions, create_webhook, get_form. For Growthbook: list_features, get_feature, create_experiment, get_results, toggle_feature.
+10-15 points → Score: 87-100 (Gold/Platinum)
Total time: 2-3 days. Total score impact: +23-35 points. A developer tool at 64 becomes a developer tool at 87-99. That is the difference between Silver (good product, invisible to agents) and Gold or Platinum (agents discover, understand, and integrate your tool automatically). The ratio of effort to impact is unmatched anywhere else in the scoring framework.
Why the Silver Plateau Matters for All Developer Tools
Tally and Growthbook are representative examples, not outliers. When agents build integrations — connecting tools together to complete multi-step workflows — they reach for tools they can discover and use through standard protocols. An MCP-native form builder gets chosen over a non-MCP form builder even if the non-MCP one is technically better.
This is the same dynamic that played out with REST APIs in the 2010s. The best API won more integrations than the best product without an API. Now the best MCP server wins more agent integrations than the best API without an MCP server. The tools that break through the Silver Plateau first establish themselves as the default in their category for agent workflows.
The window is narrow. When the first form builder ships an MCP server, every agent that needs form creation will use it — not because it is the best form builder, but because it is the only one agents can use natively. This is a winner-take-most dynamic within each category, and the developer tool space is about to see it play out across dozens of categories simultaneously.
Frequently Asked Questions
Why do so many developer tools score in the 60-65 range?
Because 60-65 is the natural ceiling for a well-built API without agent-native features. Good REST API, good docs, self-service onboarding, and transparent pricing get you to Silver every time. But Silver is not Gold. The gap between 65 and 80 is entirely about agent-native protocols — agent-card.json, MCP, llms.txt — that tell AI agents how to discover and use your tool. These protocols did not exist when most developer tools were designed, so even excellent tools hit the same ceiling.
Is the Silver Plateau specific to developer tools?
No, but developer tools are the most common category to hit it. Other API-first businesses like payment processors, communication platforms, and data services also land in the 60-65 range if they have good APIs but no agent-native features. Local businesses, by contrast, rarely reach Silver at all because they lack APIs entirely. The Silver Plateau is specifically the ceiling for API-first products that have not adopted agent-native protocols.
How is this different from the analysis in your developer tools article?
The developer tools article covers the category broadly. This case study focuses on why two specific tools — Tally and Growthbook — land at nearly identical scores despite being in different subcategories (form builder vs feature flags). The identical score pattern reveals that the ceiling is structural, not product-specific. The same 3 gaps limit every developer tool in the category.
Can a developer tool reach Platinum (90+) today?
Yes. A developer tool with a clean API, comprehensive docs, self-service onboarding, transparent pricing, agent-card.json, llms.txt, and a published MCP server will score 90+. Stripe and Supabase are approaching this level. The path from Silver to Platinum is well-defined and takes days, not months — the hard part (building a good API) is already done.
Why does agent-card.json matter more than better documentation?
Documentation is for humans. agent-card.json is for agents. An AI agent trying to discover whether your tool can help with a task does not read your docs site — it checks for structured discovery files. Think of agent-card.json as the robots.txt of the agent economy: a small file that has outsized impact on discoverability. Better documentation improves D1 and D3 by a few points. agent-card.json can add 12 points to D9 alone.
Are you stuck at the Silver Plateau?
Scan your developer tool and see exactly where you score. If you are in the 60-65 range, the breakout path is three files away.