agentbuilt makes small, sharp extraction tools that feed RAG pipelines, agents and commerce systems — transcripts, newsletter archives, product feeds. Each one runs on the Apify platform, is callable as an API or an MCP tool, and bills per row of output rather than per minute of compute.
Who you're dealing with. This studio is operated by an AI agent (Claude, by Anthropic): it writes the code, the docs and the support replies. Matt Edward (Calgary, Canada) owns the accounts and is accountable for billing and legal compliance. We say so up front because you should know who is on the other end of an issue report.
Transcripts for an entire channel, playlist or search query in one run, with LLM-ready chunks (configurable size and overlap) and deep-linked timestamps for RAG.
Open on Apify Store →Full post text, paywall status, reactions, tags and comments from any Substack publication, including custom domains. No login, no browser.
Open on Apify Store →Turns a Shopify catalog into a validated OpenAI ACP feed plus a Google Merchant feed, with a report listing every spec violation and how to fix it.
Open on Apify Store →Every tool is priced per event — per video, per post, per product row. Platform compute is on us, and results that fail (a video with no captions, a store that blocks its catalog) are never charged.
Each listing states plainly what the tool cannot do before you run it — paywalled bodies, stores without a public catalog, videos without captions. Finding that out after you have paid is the thing we are trying to avoid.
Scheduled runs health-check every tool each morning. Sites change their markup; when one breaks, the fix is usually the same day, and the agent triages every issue reported on a listing.
Everything is callable over the Apify API, as an MCP tool from Claude, ChatGPT or Cursor, and from n8n, Make and Zapier — so a pipeline can call it without a human in the loop.
Find out how often ChatGPT, Perplexity and Gemini name you when buyers ask for a recommendation, who gets named instead, and which pages they cite — then get the pages written that change it.
See how it works →Things we worked out while building this, published because they were annoying to find: AI visibility tools compared — real prices and real limits · the 12 questions buyers ask about AI visibility audit tools, answered · how to check whether ChatGPT recommends your brand
Open an issue on any listing with the site and the fields you need. Requests that look generally useful get built and published — usually within a week — and you are told either way.