Pencil
An AI design/build tool that uses six agents to craft apps in real time. It is presented as part of the emerging agentic design workflow.
Key Highlights
- Pencil is an AI design tool that uses six agents in parallel to generate app screens in real time.
- Its JSON-based pen file format links design generation to downstream web and mobile code workflows.
- Newsletter examples place Pencil in a practical stack with Claude, Claude Code, Cursor, and Expo.
- Pencil's Swarm mode was highlighted as enabling one-prompt design-to-website creation.
- For AI PMs, Pencil is most useful for rapid prototyping, workflow compression, and testing agentic collaboration patterns.
Pencil
Overview
Pencil is an AI-native design and build tool positioned within the emerging agentic design workflow. Across the newsletter mentions, it is described as using six AI agents working in parallel to craft app interfaces in real time, often inside a collaborative canvas experience. It stands out for turning a single prompt or lightweight product spec into multi-screen designs quickly, then exporting those designs as a structured JSON-based "pen file" that can feed downstream code generation and implementation workflows.For AI Product Managers, Pencil matters because it compresses the path from requirements to mockups to runnable product surfaces. In the cited examples, Pencil was used alongside Claude for requirements definition and Claude Code / Cursor / Expo for implementation, suggesting a practical AI product workflow: spec first, agentic design second, code generation third. That makes Pencil relevant not just as a design toy, but as a workflow accelerant for PMs validating concepts, aligning teams, and iterating on product UX faster.
Key Developments
- 2026-03-02: Pencil was highlighted via Tom Krcha's work as a tool that uses six AI agents to craft apps in real time, framing it as part of a broader shift toward AI-driven design workflows.
- 2026-03-08: Pencil's new swarm mode was described in detail: six AI agents powered by Cloud Opus 4.6 collaboratively designed three screens of a mobile travel log app, exported the work as a JSON-based pen file, and converted it into a React + Tailwind + Next.js website. The mention also noted that pen files can be converted to Swift iOS, Kotlin, or React Native, with community plugins for Figma and Lovable.
- 2026-03-09: Peter Yang featured Pencil's AI design tool as having a six-agent Swarm mode, a full design canvas inside Cursor and Claude Code, and a one-prompt design-to-website workflow. Two weeks after launch, the product had reportedly surpassed 100K users.
- 2026-04-02: In Peter Yang's React Native fitness app workflow, Pencil was used for AI-driven design after requirements were drafted with Claude. The end-to-end flow connected Claude for product definition, Pencil for UI generation, and Cloud Code with Expo Go for implementation and testing.
- 2026-04-02: Pencil's AI, using the Claude Opus model and six design agents, generated all UI mockups for the fitness app in under five minutes and output a `fitness.pen` JSON file, demonstrating speed and interoperability in a real build process.
Relevance to AI PMs
- Prototype faster from specs: PMs can use Pencil after drafting a lightweight PRD or `spec.md` to generate screens quickly, making it easier to validate flows, identify UX gaps, and align stakeholders before engineering starts.
- Operationalize design-to-code workflows: Pencil appears to fit into a toolchain with Claude Code, Cursor, Expo, and web/mobile frameworks, which is useful for PMs designing repeatable workflows from requirements to implementation.
- Test agentic collaboration patterns: Because Pencil uses multiple agents in parallel, PMs can study how subagents split work across screens or tasks, which is directly relevant when designing broader multi-agent product workflows.
Related
- Claude Opus model / cloud-opus-46: Pencil's swarm workflows were described as being powered by Claude Opus / Cloud Opus 4.6, indicating the model layer behind its multi-agent design generation.
- Claude Code: Frequently appears alongside Pencil as the coding layer that turns designs and specs into working applications.
- Cursor: Pencil was mentioned as having a full design canvas in Cursor, connecting AI design directly to developer workflows.
- Peter Yang: A key amplifier of Pencil in the newsletter, documenting how it fits into practical AI app-building workflows.
- Tom Krcha: Associated with Pencil's early framing as a six-agent, real-time app creation tool.
- fitnesspen: Refers to the `fitness.pen` JSON output generated by Pencil for a React Native fitness app mockup workflow.
- Expo: Part of the downstream implementation stack used after Pencil-generated designs were handed off to code.
- Lovable: Pencil's pen files were noted as having community plugin support for export to Lovable, linking it to adjacent AI app-building ecosystems.
- AI agents: Pencil is a concrete example of agentic product creation, where multiple agents collaborate on a shared output rather than relying on a single linear generation step.
Newsletter Mentions (5)
“Generated all UI mockups in under five minutes with Pencil's AI (using Claude Opus model and six design agents) outputting a fitness.pen JSON file”
Anthropic Demos Claude Code for Mobile Apps #1 ▶️ Full Tutorial: Build a Beautiful Mobile App with Claude Code in 16 Minutes Peter Yang Builds a React Native fitness tracking app in roughly two hours using Claude for requirements, Pencil for AI-driven design, and Cloud Code with Expo Go for implementation and testing. Co-created a spec.md with Claude defining three screens (add/edit workouts, live workout session, and calendar), progressive overload rules, pound/kg toggle, and dark-only theme Generated all UI mockups in under five minutes with Pencil's AI (using Claude Opus model and six design agents) outputting a fitness.pen JSON file Implemented the app in Cloud Code over three milestones, downgraded from Expo SDK 55 to SDK 54 for Expo Go on iPhone, and committed 6,400 lines of code across eight screens
“Peter Yang Builds a React Native fitness tracking app in roughly two hours using Claude for requirements, Pencil for AI-driven design, and Cloud Code with Expo Go for implementation and testing.”
Anthropic Demos Claude Code for Mobile Apps #1 ▶️ Full Tutorial: Build a Beautiful Mobile App with Claude Code in 16 Minutes Peter Yang Builds a React Native fitness tracking app in roughly two hours using Claude for requirements, Pencil for AI-driven design, and Cloud Code with Expo Go for implementation and testing. Co-created a spec.md with Claude defining three screens (add/edit workouts, live workout session, and calendar), progressive overload rules, pound/kg toggle, and dark-only theme Generated all UI mockups in under five minutes with Pencil's AI (using Claude Opus model and six design agents) outputting a fitness.pen JSON file Implemented the app in Cloud Code over three milestones, downgraded from Expo SDK 55 to SDK 54 for Expo Go on iPhone, and committed 6,400 lines of code across eight screens
“#9 in Peter Yang unveils Pencil’s AI design tool with 6-agent Swarm mode, a full design canvas in Cursor & Claude Code, and one-prompt design-to-website.”
#9 in Peter Yang unveils Pencil’s AI design tool with 6-agent Swarm mode, a full design canvas in Cursor & Claude Code, and one-prompt design-to-website. Two weeks post‐launch it’s surpassed 100K users, underscoring that craft and care still beat linear workflows.
“Six AI agents powered by Cloud Opus 4.6 in Pencil’s new swarm mode collaboratively design three screens of a mobile travel log app with Oceanania imagery and export the result as a JSON “pen file” that is then converted into a React + Tailwind + Next.js website running on port 8080.”
Six AI agents powered by Cloud Opus 4.6 in Pencil’s new swarm mode collaboratively design three screens of a mobile travel log app with Oceanania imagery and export the result as a JSON “pen file” that is then converted into a React + Tailwind + Next.js website running on port 8080. Pencil’s swarm mode (released Tuesday) assigns six subagents to design three app screens in parallel, each subagent indicated by its own cursor on the canvas. The design is stored in a JSON-based “pen file” format that can be converted to Swift iOS, Kotlin or React Native and has community plugins to export to Figma and Lovable.
“Tom Krcha’s Pencil uses six AI agents to craft apps in real time, Felix Lee leverages Figma MCP for seamless design-to-code loops, and Dylan Field redefines taste, craft, and point of view in an era of limitless ...”
#8 in Peter Yang spotlights how AI agents are revolutionizing design: Tom Krcha’s Pencil uses six AI agents to craft apps in real time, Felix Lee leverages Figma MCP for seamless design-to-code loops, and Dylan Field redefines taste, craft, and point of view in an era of limitless ... #9 𝕏 Santiago highlights that AI model access now costs just $200/month—a level of affordability unheard of three years ago—and warns this cheap pricing is built on borrowed time. He hopes token prices stay low before current subsidies dry up.
Related
Anthropic's coding-focused agentic tool for building and automating software workflows. In this newsletter it is discussed as being integrated with Vercel AI Gateway and as a Chrome extension for browser automation.
An AI coding assistant/editor that can use dynamic context across models and MCP servers to reduce token usage. Useful for AI PMs thinking about agentic workflows, context management, and efficiency.
A writer/observer mentioned for a post about how vibe coding is reshaping developer workflows. Relevant to AI PMs for workflow and interface trends.
Autonomous or semi-autonomous systems used here in sales and coding workflows. The newsletter highlights their role in replacing human SDR tasks and orchestrating complex tasks.
A no-code AI app builder referenced here as the platform used to build a production-grade SaaS product. For PMs, it illustrates how agentic coding is changing build-vs-buy and software creation economics.
Stay updated on Pencil
Get curated AI PM insights delivered daily — covering this and 1,000+ other sources.
Subscribe Free