Lenny’s Reads: “ChatGPT Apps Are About to Be the Next Big Distribution Channel: Here’s How to Build One”

Host: Lenny Rachitsky
Episode Date: January 20, 2026
Author & Main Contributor: Colin Matthews (narrated by Lenny)

Episode Overview

In this special audio edition of Lenny’s Newsletter, Lenny Rachitsky narrates a post by repeat collaborator Colin Matthews about the rapid rise of ChatGPT apps as a major new distribution channel. The episode explores why leading companies are betting on in-chat applications, how the new Model Context Protocol (MCP) powers these integrations, and delivers a step-by-step guide to building and launching your first ChatGPT app. The episode is geared toward both product leaders seeking growth opportunities and individual builders exploring micro-app creation.

Key Discussion Points & Insights

1. The New Era of In-Chat Transactions (00:55)

• Current State:
  A few months ago, ChatGPT would provide suggestions (e.g., flight info) but send users elsewhere. Now, interactive widgets from third-party apps can appear directly in chat, enabling actions like booking flights without leaving the interface.

• Distribution Opportunity:

  “ChatGPT apps represent a rare distribution opportunity, the kind that comes around once or twice a decade. The last comparable moments were the app store in 2008, the rise of SEO in the early 2000s, and maybe Shopify’s app ecosystem.”
  (Colin Matthews, 03:00)

• Major Partners:
  Early integrations include Adobe, DoorDash, Canva, Figma, Booking.com, Coursera, Expedia, Spotify, Zillow, and more, reflecting major industry adoption.

2. How Users Discover and Use ChatGPT Apps (04:00)

• Contextual Surfacing:
  Unlike traditional app stores requiring manual discovery and installation, ChatGPT can automatically surface relevant apps in-context based on conversation.

  “A user doesn’t need to know what apps are available. The model matches their intent to relevant tools.”
  (Colin Matthews, 05:10)

• App Formats:

  • Inline Mode: Cards/lists in the chat flow (great for product listings or search results)
  • Full Screen Mode: Takes over the chat UI for complex tasks (e.g., maps, design tools)
  • Picture-in-Picture: Small floating window for ongoing, peripheral tasks (music, timers)

• UX Constraint:
  Only one widget can render per message—so even if users request multiple tools, they enter tasks sequentially.

3. Architecture of a ChatGPT App (08:00)

• Three Components:

  1. Conversation: User requests are analyzed in ChatGPT; AI decides if an app/tool is needed.
  2. Tools: Defined backend functions (APIs) that describe what the app can do (e.g., “Search Restaurants”).
  3. Widgets: User-facing UIs rendered in chat, built with web tech (usually React), securely sandboxed.

• How a Tool Works:
  The app exposes functions and clear descriptions. ChatGPT matches the user’s intent, calls the right tool with parameters, and renders results via widgets.
  This creates an “orchestration loop”:

  1. User sends request
  2. ChatGPT chooses tool
  3. Widget displays result
  4. User interacts, triggering more tool calls

• Key Insight:

  “ChatGPT orchestrates the whole thing. It decides when to call tools, what parameters to pass, and how to respond to user actions. Your app just exposes capabilities and renders UI.”
  (Colin Matthews, 12:15)

4. MCP: The Technology Backbone (13:10)

• Model Context Protocol (MCP):
  A new, generalized protocol for connecting AI assistants to apps and interactive UIs, initially built by Anthropic (Nov 2024), with OpenAI rolling out support in March 2025.

  “MCP is very similar to APIs...but rebuilt for AI agents. It provides a universal way to connect apps to AI assistants.”
  (Colin Matthews, 14:30)

• Standardization:
  OpenAI and Anthropic now collaborate on MCP apps, creating a universal ecosystem.

• Relation to AI Agents:
  Like AI agents, but ChatGPT retains orchestration control (not the app builder).

5. Creating Your First ChatGPT App – Step-By-Step Guide (16:05)

• Importance of Tool Naming & Descriptions:
  This affects how the AI “finds” and recommends your app, much like SEO for web.

• Two Methods Explained:

  Option 1: Replit (18:05)

  • Import example apps (official OpenAI examples or Colin’s starter project).
  • Set up in Replit (5-10 min suggested).
  • After running, connect to ChatGPT via MCP URL.
  • Practical advice:

    “Most Vibe coding tools aren’t built to help you understand MCP and ChatGPT apps...That’s why I decided to build one myself as a second option.”
    (Colin Matthews, 21:00)

  Option 2: Chippy (23:05)

  • A new AI agent purpose-built for prototyping ChatGPT apps.
  • Features: preview/testing, generate specs, no deployment headache, free to start.
  • Example built: Maven Lightning Lesson Finder with video playback (inline & picture-in-picture modes).
  • Step-by-step: plan via Chippy, implement tool, preview UI, connect via MCP, test in ChatGPT.

• Iterate and Expand:
  Once working, enhance your app with features like authentication, live data, and multi-tool flows.

• Beyond Basics:

  “You can build full complex applications directly into ChatGPT.”
  (Colin Matthews, 28:00)

  • Colin shares building a Dungeon Explorer game via Chippy—demonstrating depth and persistence in-app context.

Notable Quotes & Memorable Moments

• On Big Picture Opportunity:

  “When a new distribution channel opens up at scale, the companies that move early capture habits that are hard to break later.”
  (Colin Matthews, 02:30)

• On Contextual UX:

  “Ask about travel plans and Expedia appears. Mention that you need a design and Canva surfaces. Ask about ordering groceries and an instacart cart shows up.”
  (Colin Matthews, 05:10)

• On AI as Orchestrator:

  “The key insight is that ChatGPT orchestrates the whole thing. ... Your app just exposes capabilities and renders UI.”
  (Colin Matthews, 12:15)

• On Builder Tools:

  “Most Vibe coding tools aren’t built to help you understand MCP and ChatGPT apps.” (Colin Matthews, 21:00)

Important Timestamps

• 00:55 – How ChatGPT switched from “advisor” to transaction platform
• 03:00 – The size and rarity of this distribution opportunity
• 05:10 – Contextual surfacing of apps within ChatGPT
• 08:00 – Architecture of a ChatGPT app explained
• 12:15 – ChatGPT as orchestrator and the loop of interaction
• 14:30 – Introduction to the Model Context Protocol (MCP)
• 16:05 – How to make your app “discoverable” (answer engine optimization)
• 18:05 – Replit setup walkthrough
• 23:05 – Building and previewing with Chippy
• 28:00 – Building complex apps/games as potential use cases

Final Takeaway

The episode urges listeners to capitalize on ChatGPT’s emerging app ecosystem—comparing the opportunity to historic inflection points like the launch of the app store. With tools like MCP, and prototyping platforms such as Chippy, both major companies and solo founders have a chance to create remarkably interactive in-chat app experiences at unprecedented scale. Listeners are encouraged to get started immediately to capture user habits early in this new distribution wave.

(For visuals & code examples, Lenny and Colin recommend visiting the written version linked in the show notes.)