What AI agents actually see when they search for your Apify Actor

Description and use

The author investigated how AI agents discover new pay‑per‑event Actors in the Apify Store via the MCP server, showing that only Actors with limited permissions and no standby mode are visible, and identified two pitfalls in measurement.

Problem

Publishers want to know how AI agents discover new pay‑per‑event Actors in the Apify Store, especially for cold‑start Actors with no reviews or users.

Outcome

The MCP search pipeline exposes only free or pay‑per‑event Actors with limited permissions and no standby mode. The author’s three new Actors were discoverable by AI agents, but their ranking depended on keyword relevance. Two measurement pitfalls were identified: mis‑extracting the session‑id and mis‑parsing the SSE stream.

Workflow

• 1. Export APIFY_TOKEN and initialize a session with a POST to https://mcp.apify.com/, capturing the mcp‑session‑id from the response header.
• 2. Call the search‑actors tool via JSON‑RPC over curl, passing keywords, limit, and offset, and include the session‑id header.
• 3. Parse the Server‑Sent Events stream to extract the JSON payload, then iterate over offsets to page through results up to the desired depth.
• 4. Record the positions of the target Actors in the search results and compare them to the human‑facing Store search results.

Evidence

The server returns only free and pay‑per‑event Actors to agentic callers — rental‑model Actors are excluded

Source

I wanted the rawest possible view — no SDK, no client‑side magic — so I spoke JSON‑RPC to the server directly with curl

Source