ExtensionDash: Tracking a Browser Extension via AI Agent
Developer launches ExtensionDash to scrape store ranks and ratings, offering Model Context Protocol support for AI agents to query insights directly.
Stock photo for illustration only, not from the actual event
- Developer launches ExtensionDash to solve the lack of search position tracking on Chrome Web Store and Firefox.
- System scrapes stores nightly, recording installs, ratings, category ranks, keyword positions, and competitors.
- Connects via Model Context Protocol (MCP) with 15 tools, enabling AI agents to read and analyze data instantly.
- Available for free with limits of 3 extensions per account and 15 keywords per extension, requiring no credit card.
After shipping two browser extensions this year — Reverse Image Search Anywhere and Stackpeek — a developer found a frustrating gap: neither the Chrome Web Store nor Firefox provides the most critical metric for visibility: where an extension appears when users type actual search keywords. Answering whether a title change worked required opening incognito windows, typing keywords, and counting results manually in three languages.
To fix this, the developer built ExtensionDash, a system that nightly scrapes both stores and records install counts, ratings, category ranks, keyword positions, competitor data, and listing copy while maintaining a complete history. Rather than treating an AI agent merely as an export button, the system treats the agent as a first-class reader.
The platform exposes a POST /mcp endpoint implementing the Model Context Protocol over stateless HTTP, authenticated via a per-user token. It provides 15 tools total — 11 read tools and 4 write tools that cannot delete historical data. Everything rendered on the dashboard can be directly read and interpreted by an AI agent.
Stock photo for illustration only, not from the actual event
The Model Context Protocol (MCP), created by Anthropic, establishes an open standard for securely connecting AI models to external data sources and developer tools. By exposing ExtensionDash over an HTTP MCP transport, developers allow external AI assistants to query deep store analytics programmatically without writing custom integration layers.
A core design challenge involved making data absence legible to an agent. While human users interpret blank table cells using context, an agent receiving a JSON document containing a null value might misinterpret it as zero. To prevent tools from confusing "no ranking for this keyword" with "not yet scanned," the platform explicitly separates three distinct states: never scanned, scanned and found, and scanned but absent.
Furthermore, every position record travels alongside its scan depth. Ranking #1 out of 9 results differs fundamentally from ranking #1 out of 50. Recording keyword positions across all 50 scanned results prevents arithmetic misrepresentation, while adding a competitor automatically backfills rank history to the day tracking began.
The tech stack remains intentionally straightforward: Rails 8, Postgres, Solid Queue running browser tasks on a single-threaded queue, Nokogiri and Ferrum for Chrome automation, the public AMO API for Firefox, and Kamal for deployment straight from a laptop.
"Connect it to your agent and ask it something real... tell me where it stalls, what it guesses at, and what it reaches for that is not there."
Anton Kopylov
ExtensionDash is currently open, free, and available at extensiondash.com without requiring a credit card. Usage limits cap accounts at 3 extensions and 15 keywords per extension because each keyword triggers a nightly browser session in perpetuity. Users can test the MCP integration and share feedback with the creator at anton@kopylov.net.
Source: Dev.to
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment