LoginSign Up
← Back to all sparks
A

Apify

ANALYTICS
Velocity6.3

Apify is rebuilding the Actor platform as MCP-first agent infrastructure.

mcpai-agentsmarketplace-discoveryapipermissionsstandby-actors
◆Current state
Apify's Actor platform is reorienting around AI agents. Recent releases add MCP connectors for authenticated apps, a redesigned MCP configurator spanning major LLM clients, interactive OpenAPI endpoints for standby Actors, and stricter permission defaults framed explicitly around agent safety. The marketplace itself is gaining agent- and search-readable surfaces.
◆Where it's heading
The throughline is making Actors first-class tools for LLM agents: callable, documented, permissioned, and discoverable. OpenAPI docs and the configurator lower the friction of letting an agent invoke an Actor it didn't write, while permission gates add a safety counterweight. Discovery features extend the same agent-centric logic to distribution on Apify Store.
◆Prediction
Expect broader MCP coverage — more Actors marked MCP-compatible and tighter authenticated connector flows — alongside further agent-oriented discovery surfaces on the Store.

◆Recent moves

  1. 2d ago

    New Actor creation flow

  2. 3d ago

    Publish tasks for your Actor to get more users

    Published tasks give each Actor task its own search-indexed Store page, extending Apify's discovery surface from the Actor itself down to specific, long-tail use cases. It is a distribution play aimed at both search engines and AI agents, consistent with the platform's agent-readable direction.

  3. 18d ago

    MCP connectors are live. Actors now work where you do.

    ⚡ SPARK

    MCP connectors close the biggest gap in Apify's agent story: Actors could only touch the open web, but can now act inside authenticated apps like Notion, Slack, and GitHub without ever seeing credentials. This is the load-bearing piece of the agent-infrastructure arc.

  4. 1mo ago

    Interactive OpenAPI documentation for standby Actors

    Standby Actors that ship an OpenAPI spec now expose an Endpoints tab with live, authenticated request execution in the browser. It is part of making Actors legible and callable — the same instinct behind the MCP work, applied to documentation.

    View source ↗
  5. 1mo ago

    Full-permission Actors now require approval

    A one-time confirmation gate for full-permission Actors tightens defaults across Console, API, CLI, schedules, and webhooks. Framed around least privilege for users and AI agents, it is the security counterweight to Apify's push to let agents invoke unfamiliar Actors.

    View source ↗
  6. 2mo ago

    Multiple datasets for Actors

    Actors can now write to multiple schema-validated datasets per run, letting creators separate clean output from debug data. A structural improvement to how Actor output is organized rather than a directional shift.