Selling SDKs in the era of many Claudes (Interview)

What It Covers

Steve Ruiz, founder of TLDraw, joins The Changelog to discuss building and selling an SDK-based infinite canvas business with roughly 20 employees, navigating commercial licensing from MIT to a license-key enforcement model, generating nearly $1M in year-one revenue, and how agentic AI is reshaping both his development workflow and the demand for self-hostable, internally owned software tooling.

Key Questions Answered

• •SDK Licensing Progression: TLDraw cycled through three distinct license models before finding traction. Starting MIT-licensed, then noncommercial-only, then a watermark-removal model, and finally a production license-key enforcement model where the SDK disappears without a valid key. The watermark model actively suppressed deal sizes by anchoring customer perception of value to removing a small SVG rather than to the canvas's actual engineering depth. Switching to license-key enforcement removed that anchor entirely.
• •Year-One Revenue Benchmark: With a bare-bones go-to-market structure, no formal sales team, and fully negotiated individual deals, TLDraw generated close to $1M in its first commercial year. The approach was deliberately manual: negotiate every contract, test price elasticity by raising the ask when customers agreed quickly, and use each deal as a discovery session to understand how deeply the SDK embedded into a customer's product and revenue stream.
• •Scope-Limited Enterprise Deals: When large enterprises want TLDraw but only for one team or one feature, limiting the license agreement to that specific application or product line allows a lower price point without discounting the full company license. This creates a land-and-expand motion where adjacent teams later request their own agreements, effectively growing revenue inside the account without requiring a renegotiated enterprise-wide contract upfront.
• •Agentic AI Compresses Feature Cycles: A multi-week engineering project — a Comfy UI-style async image pipeline starter kit — was taken from concept to 80% completion in roughly two hours by handing Claude a description and links to reference products. The remaining work was UX steering and architecture review, not code writing. The full cycle from idea to shipped starter kit ran four to five days versus the anticipated two-engineer, multi-week timeline.
• •Internal Tooling as a New Revenue Category: Enterprise customers are now approaching TLDraw not for GDPR data-sovereignty reasons but because external SaaS tools expose insufficient API surfaces for AI agents. When a SaaS product's API omits data — such as Notion's version history being inaccessible programmatically — agents cannot function effectively. Customers want source-owned, internally deployable software they can extend freely, creating a new inbound category distinct from traditional self-hosted or SDK licensing.
• •License-Key Enforcement Enables Sales Intelligence: Requiring a license key for production use, while offering free keys for noncommercial and hobby projects plus a 100-day commercial trial key, generates a structured lead pipeline. Developers who register keys reveal their domain and email, allowing the sales team to identify the likely project owner or budget holder within the organization and initiate outreach — replacing the prior approach of monitoring CDN logs for watermark SVG requests.
• •Performance Optimization as Differentiation: TLDraw targets 120 frames per second on a React DOM canvas — no Canvas element, just divs and SVGs — through specific micro-optimizations: replacing Lodash set comparisons that silently convert sets to arrays, encoding draw-shape point data as 16-bit floats in fixed-size binary rather than JSON objects, and skipping hover-detection during camera panning on large boards. That last change alone halved frame duration during scroll on dense canvases.