My Actors worked fine until an AI agent called them
On 27 July I published a scraper that pulls product cards and specifications from online stores. It was my fourteenth Actor. The others cover company due diligence by tax ID, government tender feeds, cadastral records, marketplace seller leads. All of them had months of green runs behind them. Then I connected the Apify MCP server to Claude. It exposes Actors as tools an AI client can call over…
On July 27, the author published a scraper that pulls product cards and specifications from online stores. This was the fourteenth Actor they created, all of which had previously operated without issue. The author connected their Apify MCP server to Claude, an AI agent, enabling Actors to be called as tools via the Model Context Protocol.
They provided the agent with a task in plain language: find a product across those stores and return the specifications. However, the agent successfully ran the scraper but returned zero results despite no error being reported. This occurred because the agent had read the author's input schema as if they had the target website open, a task that a human could easily recover from.
The author identified two key issues: input fields and billing, which caused financial losses. The changes they made to address these issues involved moving the resolution of internal option ids and opaque integers to the Actor code, using enums with readable titles instead of free-form strings, and adding a reason for records that could not be found.
The author implemented these changes to prevent agents from guessing based on incomplete information and to provide more informative results.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
