Web Development ChatGPT Plugins
Last reviewed
Sources
17 citations
Review status
Source-backed
Revision
v3 ยท 2,721 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Sources
17 citations
Review status
Source-backed
Revision
v3 ยท 2,721 words
Add missing citations, update stale details, or suggest a clearer explanation.
See also: ChatGPT Plugins, ChatGPT Plugin Categories and Web Development
Web Development ChatGPT Plugins were a now-retired category of third-party extensions for ChatGPT that let the model perform tasks adjacent to building, browsing, scraping, and summarizing the public web. The category, like the entire plugin platform, is deprecated: OpenAI stopped allowing new plugin conversations on March 19, 2024, and shut down all remaining plugin-based chats on April 9, 2024, redirecting developers and users to GPTs (custom versions of ChatGPT) and GPT Actions instead.[2][3] These web tools should be treated as historical: they no longer run inside ChatGPT, and the workflows they pioneered are now handled by built-in browsing, Custom GPTs in the GPT Store, GPT Actions, and the Model Context Protocol.
During their roughly twelve-month run, these plugins sat at the intersection of AI assistance and front-end work, letting GPT-4 reach beyond its training cutoff to fetch live pages, extract structured data, render diagrams, and in some cases scaffold simple websites. After the deprecation, the surviving tools migrated to one of three replacement environments: a Custom GPT in the GPT Store, a standalone Perplexity or ChatGPT feature, or a third-party browser extension and IDE integration such as those offered by Cursor.
OpenAI shipped the plugin platform on March 23, 2023, as an experimental, limited-alpha way to extend ChatGPT Plus with external services.[1] Plugins ran as small OpenAPI specifications that the model could call mid-conversation. The first wave was built with launch partners including Expedia, FiscalNote, Instacart, KAYAK, Klarna, Milo, OpenTable, Shopify, Slack, Speak, Wolfram, and Zapier, while OpenAI hosted two of its own: a web browser and a code interpreter.[1]
The catalog grew quickly. By late 2023 it spanned travel, shopping, productivity, education, finance, and developer tooling, reaching roughly 800 plugins by August 2023 and 1,033 by December 26, 2023.[14] Web development plugins were among the earliest and most-used because the base model could not, on its own, fetch URLs, run searches, or interact with live pages.
The "Web Development" tag inside the ChatGPT plugin store was always a loose grouping. It collected tools that touched the web in some way: browsing helpers, scrapers, link readers, diagram generators, page builders, and developer-oriented utilities. Many of those tools were also surfaced under adjacent tags such as research, productivity, or design, so the boundaries blurred.
Usage data shared by OpenAI showed that adoption of plugins remained concentrated among a small group of power users, and most ChatGPT Plus subscribers either never enabled them or used only one or two consistently. On March 19, 2024, OpenAI announced that no new conversations could be started with plugins, and on April 9, 2024, all remaining plugin-based chats were shut down.[2][3] From the platform launch on March 23, 2023, to the sunset on April 9, 2024, the entire era lasted just over twelve months.
The replacement path was GPTs, also called Custom GPTs, which OpenAI had introduced at its first DevDay in November 2023 and distributed through the GPT Store starting January 10, 2024. At the store's launch OpenAI said users had already created more than 3 million custom GPTs, a scale the plugin catalog never approached.[15] GPTs expose a similar action mechanism but with a guided builder, a public store, and tighter integration into the main ChatGPT interface.
The web development tag covered four overlapping use cases:
| Use case | What the plugin did | Typical example |
|---|---|---|
| Web browsing | Fetched the live contents of one or more URLs and returned text, links, or images to the model | Browse with Bing, WebPilot, BrowserOp |
| Link reading and summarization | Took a single URL and produced a summary, transcript, or structured extract | Link Reader, URL Reader |
| Web scraping | Pulled structured data such as tables, prices, or contact details from arbitrary pages | Scraper, GPT Scraper variants |
| Site generation and design | Produced HTML, CSS, or hosted pages from a natural-language brief, sometimes with image generation | DesignerGPT |
| Developer visualization | Rendered diagrams, flowcharts, or wireframes inline in the chat | Show Me Diagrams |
A few plugins blurred the line further by combining browsing with extra services such as text rewrites or translation. The first-party browsing plugin maintained by OpenAI itself was the most important entry in the category because every other web tool either competed with it or filled a gap it could not cover.
Browse with Bing was OpenAI's own web browsing plugin, sometimes referred to internally as the "web browser" plugin. It gave the language model access to a text-based browser that issued GET requests and returned the resulting page text. The design intentionally excluded form submissions and other write operations to limit safety risks. OpenAI said it built the browser "with safety in mind" and designed it "to operate as a good citizen of the web," using the user-agent token ChatGPT-User, honoring sites' robots.txt files, applying rate limits, and citing the sources it visited so that "this gives users transparency and helps content creators get the credit they deserve."[1]
The feature first appeared in May 2023 for ChatGPT Plus subscribers, was disabled briefly on July 3, 2023, after users found ways to make it bypass paywalls and reproduce full article text, and was later restored with stricter content policies.[5][6] The functionality was eventually folded into ChatGPT's core product as the built-in browsing tool, which removed the need for a separate plugin.
WebPilot was one of the most popular third-party browsing plugins. It accepted one or more URLs and let the user ask for an extract, a rewrite, a translation, or a structured summary of the contents. The plugin became known partly because of its public backstory: the team behind it open-sourced an early version on GitHub and discussed the build process on the OpenAI developer forums.[9] WebPilot remained available as a Chrome and Edge extension after the plugin platform was retired, and the team also published a Custom GPT version on the GPT Store.[10]
BrowserOp, originally released as BrowserPilot before being renamed, was marketed as an efficient way to search and read across many web pages in a single ChatGPT turn. Given a URL, it would fetch and parse the page; given a free-text query, it would treat it as a search request and return real-time results. BrowserOp also offered text rewrites and translations, putting it in roughly the same niche as WebPilot.[8] After plugin deprecation it relaunched as a Custom GPT.
Link Reader was a single-URL summarization plugin that worked across a wider range of formats than most competitors. It read web pages, PDF documents, PowerPoint files, Word documents, and image files, and could transcribe YouTube videos via captions. Typical prompts looked like "summarize the content of this link" followed by a URL.[13] Link Reader was removed from the plugin store before the broader April 2024 shutdown.
Several plugins under names such as Scraper and GPT Scraper exposed simple HTML scraping endpoints to ChatGPT. The most common workflow was to pass a URL and ask the model to pull a list of products, headlines, links, prices, or images from the rendered page. These tools overlapped with general-purpose scraping APIs from companies such as Apify, Bright Data, and Oxylabs, several of which published their own ChatGPT plugin wrappers. Most of these now exist as Custom GPTs or stand-alone APIs.
Show Me Diagrams, sometimes shortened to Show Me, let ChatGPT render flowcharts, mind maps, sequence diagrams, organizational charts, and architecture sketches inline. Without the plugin the model would emit raw Mermaid or PlantUML source code that the user had to copy elsewhere to render. With the plugin installed, the diagram appeared as an image in the chat.[11] Show Me was popular among software engineers using ChatGPT for system design exercises and documentation drafts. It survived the plugin shutdown as a Custom GPT.
DesignerGPT generated and hosted small static websites from a natural-language description. It paired with DALL-E to insert generated imagery and could iterate on background colors, fonts, headings, and layout via follow-up prompts.[12] The output was a hosted page with a public URL, useful for portfolios, landing pages, and quick prototypes. DesignerGPT was rebuilt as a Custom GPT after the plugin platform was retired.
| Task | Plugin pattern | Typical replacement after April 2024 |
|---|---|---|
| Summarize a long blog post | Pass URL to Link Reader or WebPilot | Built-in browsing in ChatGPT, Perplexity, or a Custom GPT |
| Extract a list of items from a page | Pass URL to a Scraper plugin | Custom GPT with an Action against a scraping API |
| Compare prices across several pages | Pass URLs to BrowserOp | ChatGPT browsing or Perplexity Pro Search |
| Sketch a system architecture diagram | Ask Show Me Diagrams for a Mermaid render | Show Me Custom GPT, or local Mermaid Live Editor |
| Build a single-page portfolio site | Brief DesignerGPT and iterate | DesignerGPT Custom GPT, Cursor, or hand-written HTML |
| Read up-to-date documentation for a library | Pass URL to Browse with Bing or WebPilot | Built-in ChatGPT browsing or Cursor's docs feature |
These workflows show why the category was useful: the base GPT-4 model had a fixed training cutoff, so any task that depended on recent web content needed an external fetch step. Plugins were the simplest way to provide that step before browsing was added directly to the product.
The public reasons OpenAI gave for sunsetting plugins applied with extra force to web development tools. Browsing, link reading, and scraping were the use cases most likely to overlap with built-in features that OpenAI was already shipping or planning to ship. Three pressures pushed the category toward retirement:
By April 9, 2024, every plugin-based chat was disabled.[2] Existing plugins either disappeared or rebuilt themselves as Custom GPTs, browser extensions, or standalone services.
The gap left by web development plugins was filled by several different tools rather than a single replacement.
ChatGPT shipped built-in browsing as a default capability for paid users, removing the need for a separate plugin in most situations. The browsing tool retrieves pages on demand, cites sources, and integrates with the main chat history. ChatGPT Search, announced in late 2024, extended this further with a search-engine-style interface that returned summarized answers with linked sources.
The most direct migration path. Show Me Diagrams, DesignerGPT, WebPilot, BrowserOp, and many scrapers re-launched as GPTs in the GPT Store. The interaction model is similar to the old plugins, but discovery happens through the store rather than a per-conversation toggle.
GPT Actions are the successor to the OpenAPI mechanism that plugins used. A builder attaches an OpenAPI schema to a custom GPT so the model can call an external RESTful API in natural language, which is essentially the plugin contract rebuilt inside the GPTs framework. This is why most plugin developers could port their backend to a Custom GPT with little change.[3]
The Model Context Protocol (MCP), an open standard introduced by Anthropic in November 2024 for connecting AI models to external tools and data, has become the broader successor pattern. OpenAI added MCP support to its Agents SDK in March 2025, and in 2025 to 2026 the company rebranded ChatGPT's third-party integrations: on December 17, 2025, OpenAI renamed "connectors" to "apps," a unified surface that covers both interactive app UIs and the connectors that let ChatGPT search and reference external data.[16][17]
Perplexity emerged during the plugin era as a search-first AI assistant and grew quickly afterward. It offers free conversational search, a Pro tier with deeper sourcing, and a focus mode for academic or social-media-only results. Many users who relied on WebPilot or Link Reader for quick page summarization moved to Perplexity because the entire product is built around that workflow. Anthropic added web search to Claude in 2025, giving users another high-quality option for browse-and-summarize tasks; Claude's long context window made it well-suited to large pages or batches of URLs.
For coding-adjacent web tasks such as scaffolding pages, reading documentation, or refactoring CSS, AI-native editors like Cursor absorbed much of the workload. They combine in-IDE browsing, retrieval over the project, and inline edits in a way that browser-based plugins could not match.
Web development overlapped substantially with several adjacent tags in the ChatGPT Plugin Categories taxonomy:
| Adjacent category | Overlap |
|---|---|
| Research | Both relied heavily on browsing and link summarization |
| Productivity | Diagram and site-generation tools were also tagged here |
| SEO and marketing | Scraping and site-audit plugins crossed both tags |
| Developer tools | Show Me Diagrams and code-search plugins appeared in both |
Because plugins could be tagged with multiple categories, the same tool often appeared in two or three places. The full inventory is documented at List of ChatGPT Plugins.
The web development plugins are now a closed chapter, but several patterns they popularized survived. Inline browsing, URL summarization, web-augmented prompts, and chat-driven site generation are all standard features of modern AI assistants. Tools that began as plugins, including WebPilot, BrowserOp, Show Me Diagrams, and DesignerGPT, are still operating in 2026 under the Custom GPT or browser-extension model. The original plugin store itself is read-only and exists mainly as a historical reference.
For researchers studying the early period of AI tool ecosystems, the category is a useful case study in how a vendor-led platform consolidates third-party extensions into native features once usage patterns become clear. It also shows the pace of the change: from plugin launch in March 2023 to platform sunset in April 2024 was just over twelve months.