Why Craft.do Scores 68: The Note-Taking App That Outscores Most Enterprise SaaS
Craft.do is a document and note-taking app. It scored 68 Silver on the Agent Readiness Score — tied with Stripe and Slack. That is not a typo. A focused note-taking app with a clean API outscores most enterprise SaaS platforms with 100x the engineering headcount. Here is why, and what pattern it reveals about agent readiness.
The Full Scorecard: 68 Silver
Craft.do scored 68 out of 100, placing it in the Silver tier (60-74). Here is the dimension-by-dimension breakdown showing where it excels and where the gaps are.
D1 Discovery
62/100Craft has a public developer documentation site, OpenAPI-style endpoint references, and is listed on integration directories. Missing: agent-card.json, llms.txt. Agents can find the API through standard web search but cannot auto-discover capabilities through agent-native protocols.
D2 API Quality
78/100Clean REST API with structured JSON responses. Document, block, and folder operations are well-typed with consistent naming. Error responses include error codes and messages. Versioned endpoints. This is Craft's strongest dimension — the API feels like it was designed for programmatic use, not bolted on.
D3 Onboarding
72/100Self-service API key generation from the developer portal. Documentation includes quickstart guides and code examples. OAuth2 for user-context operations. Clear scoping of permissions per token.
D4 Pricing Transparency
55/100Craft has clear consumer pricing for personal/business plans. API access is included with paid plans but the exact rate limits and API-specific pricing are not published in machine-readable format. Agents cannot programmatically determine if a given operation will be free or paid.
D5 Payment
48/100Standard web-based payment flow. No API for subscription management, usage-based billing, or programmatic upgrades. An agent managing a user's SaaS subscriptions cannot interact with Craft's billing system.
D6 Data Quality
74/100Documents and blocks return well-structured JSON with rich metadata — creation dates, modification times, content types, nested structures. Markdown content is cleanly formatted. Search results include relevance signals.
D7 Security
75/100HTTPS enforced. OAuth2 with scoped tokens. Rate limiting with clear headers. No exposed internal errors. Security headers present. Above average for a SaaS application of this size.
D8 Reliability
71/100Fast response times (sub-200ms for most operations). Consistent uptime based on historical checks. Proper HTTP status codes. Error responses are structured and actionable. No public status page dedicated to API uptime, which would push this higher.
D9 Agent Experience
58/100The API works well for programmatic document management. However, there is no MCP server, no agent-card.json, no llms.txt, and no agent-specific tooling. An agent must discover and integrate manually rather than through standard protocols.
The Pattern: Why Single-Purpose Apps Consistently Score Silver
Craft is not an anomaly. Across our scans of 500+ businesses, single-purpose apps with well-designed APIs consistently cluster in the Silver tier. The pattern is remarkably consistent and reveals something important about what makes an API agent-ready.
Single-purpose with a clean domain model
Craft does one thing well: structured documents with blocks. The API reflects this clarity — documents, blocks, folders. No ambiguity, no feature sprawl. Single-purpose apps map naturally to clean APIs because the domain model is simple.
Also seen in: Stripe (payments), Resend (email), Supabase (database)
API designed for developers, not bolted on
Craft's API uses consistent naming, typed responses, and proper REST conventions. It feels like the API was a first-class product, not an afterthought. Companies that build their API alongside their product consistently score higher.
Also seen in: Vercel (deployments), Cloudflare (DNS/CDN), Linear (issues)
Self-service onboarding with clear documentation
API key generation, quickstart guides, code examples, and a developer portal that does not require contacting sales. This pattern lifts D3 Onboarding scores significantly and is common among developer-focused products.
Also seen in: Postmark (email), Algolia (search), Neon (database)
The inverse pattern is equally telling. Enterprise platforms that try to do everything — CRM + marketing + analytics + support — often have inconsistent APIs across modules, legacy endpoints that break conventions, and onboarding flows that require sales calls. A 500-person engineering team building 50 products scores lower than a 20-person team building one product with a clean API.
This is directly relevant to the developer tools vertical, where focused products like Resend (the only Gold-tier business) demonstrate that API quality beats feature count every time.
Three Files Standing Between Silver and Gold
Every Silver-tier business we have scanned shares the same gap: the three agent-native files that separate a good API from an agent-ready one. Craft is no exception. Its API quality is Gold-level. Its agent discoverability is not.
agent-card.json
A JSON file at /.well-known/agent-card.json that describes the API's capabilities, authentication methods, and available tools in a format agents can read automatically. Without it, agents must be explicitly configured to use Craft — they cannot discover it on their own.
llms.txt
A text file at /llms.txt that tells AI models what the service does, how to authenticate, and what endpoints are available. Functions like robots.txt but for AI agents instead of search crawlers. Takes 10 minutes to create.
MCP Server
An MCP (Model Context Protocol) server that wraps Craft's API in agent-callable tools. Tools like create_document, search_documents, update_block, list_folders. This is what transforms a good API into an agent-native service.
The Gold gap is always the same three files: Across Stripe (68), Slack (68), Craft (68), Vercel (69), and Supabase (69), every Silver-tier business is missing agent-card.json, llms.txt, and/or an MCP server. The underlying API quality is already Gold-level. The agent-native discovery layer is what is missing. This is why we built the SaaS Agent Readiness guide — to help companies close this exact gap.
What an Agent Does with Craft Today vs with an MCP Server
Imagine an AI assistant managing a knowledge worker's documents. The user says: “Summarize everything I wrote about project X this week and create a status update.”
Today (Silver, no MCP)
- 1.Agent knows Craft has an API from its training data
- 2.User must provide API credentials and configure the connection manually
- 3.Agent must be programmed with Craft-specific endpoint knowledge
- 4.Works, but requires upfront setup per user
With MCP Server (Gold)
- 1.Agent discovers Craft via agent-card.json automatically
- 2.Connects to MCP server and reads available tools
- 3.Calls search_documents, reads content, creates status update
- 4.Zero configuration needed — works out of the box for every user
The difference is not in capability — Craft's API can do everything an MCP server would expose. The difference is in discoverability and zero-configuration access. An MCP server turns Craft from “an API that agents can use if configured” to “a service that agents discover and use automatically.” That is the Silver-to-Gold leap.
Lessons for Every SaaS Company
Craft's score reveals actionable lessons for any SaaS company evaluating their agent readiness.
API quality matters more than API size
Craft has a relatively small API surface — documents, blocks, folders, search. But every endpoint is well-designed, consistently named, and properly typed. A 20-endpoint API that scores 78 on D2 API Quality beats a 200-endpoint API that scores 45 because of inconsistencies and legacy baggage.
Self-service onboarding is a prerequisite
Any API that requires contacting sales, scheduling a demo, or waiting for approval to get API keys loses D3 Onboarding points. Craft lets developers generate keys from the dashboard. This is table stakes for Silver tier.
The same three files block everyone from Gold
If your API is already good, you are likely stuck at Silver for the same reason as Craft, Stripe, and Slack: missing agent-card.json, llms.txt, and an MCP server. These are the cheapest, highest-ROI improvements available.
Focus beats feature sprawl
Craft does one thing and does it well. Its API reflects that clarity. If your product does 10 things, your API probably has inconsistencies across modules. Consider scoring each module separately and fixing the worst ones first.
Frequently Asked Questions
Why does a note-taking app score as high as Stripe?
Because agent readiness measures API quality and accessibility, not company size or revenue. Craft has a well-designed REST API with structured responses, self-service onboarding, proper authentication, and good documentation. These are the same qualities that make Stripe agent-ready. The score measures how easily an AI agent can discover, connect to, and use your service — and Craft's clean API design puts it on par with much larger platforms.
How could Craft.do reach Gold tier?
By adding three files. An agent-card.json at /.well-known/agent-card.json describes capabilities for agent discovery. An llms.txt file tells AI models how to interact with the service. An MCP server wraps the API in agent-callable tools. These three additions would lift Craft from 68 Silver to approximately 78-83, firmly in Gold territory. The API quality is already there — the gap is purely in agent-native discovery and tooling.
Is 68 Silver good?
Extremely. The average Agent Readiness Score across 500+ businesses we have scanned is 43. Only 8% of scanned businesses score Silver or above. Craft at 68 outperforms 92% of all businesses and ties with Stripe (68) and Slack (68), both companies with massive engineering teams. Silver means agents can effectively use your API with some manual configuration — it is just not auto-discoverable yet.
What makes single-purpose apps score higher?
Single-purpose apps have simpler domain models, which naturally produce cleaner APIs. A note-taking app has documents, blocks, and folders. A payment processor has charges, customers, and subscriptions. When the domain is narrow, the API surface is consistent and predictable. Enterprise platforms that do 50 things often have inconsistent APIs, legacy endpoints, and documentation gaps across their sprawling surface area.
Where does your SaaS product score?
Get your free Agent Readiness Score in 60 seconds. See your dimension-by-dimension breakdown and learn exactly which files to add to reach the next tier.