The "build and deploy MCP servers" content currently lives only on the blog (Build and deploy MCP servers in TypeScript). A Slack thread flagged that this should be promoted into docs, broadened in scope, and paired with a starter repo.
Current docs cover MCP almost entirely from the consumer side. Connecting clients to mcp.apify.com, configuring tool selection, OAuth (sources/platform/integrations/ai/mcp/). The publisher side, how to wrap an MCP server as an Actor, monetize it, register it with the Apify MCP server, is a one-liner in develop_ai_agents.md and conceptual prose in the Academy actorization playbook. No step-by-step guide exists.
Scope
A new how-to guide covering:
- Both transports as first-class paths: stdio (wrap existing stdio MCP server) and native remote streamable HTTP (not just proxy mode, which is all the current TS template covers)
- Step-by-step using existing templates: TypeScript MCP server, Python MCP server
- Configure standby mode, pay-per-event monetization, Actor metadata (webServerMcpPath, etc.)
- Explain the re-publishing model: when an MCP server Actor is in the store, its tools are exposed by mcp.apify.com - via explicit registration (?tools=user/actor-name) or dynamic discovery
- Explain the monetization narrative: any Apify user can call paid third-party MCP servers through a single Apify account; Apify handles billing and revenue share to creators
- Deploy via apify push and publish to Apify Store
Companion work
- Cross-links from:
- sources/platform/integrations/ai/mcp/index.md (consumer-side hub)
- sources/platform/actors/development/quick-start/develop_ai_agents.md (templates list)
- sources/academy/build-and-publish/actor-ideas/what_software_an_actor_can_be.md (§ MCP servers and tools for AI)
- sources/academy/build-and-publish/how-to-build/actorization_playbook.mdx (examples table)
Acceptance criteria
- Both TypeScript and Python templates referenced with deploy-ready configs
- Re-publishing via mcp.apify.com documented with a worked example (explicit and dynamic discovery)
- Monetization section explains PPE setup and revenue share for MCP-server Actors
- pnpm run build passes; pnpm run lint clean; all links resolve
- Cross-links added from the four touchpoints above
References
The "build and deploy MCP servers" content currently lives only on the blog (Build and deploy MCP servers in TypeScript). A Slack thread flagged that this should be promoted into docs, broadened in scope, and paired with a starter repo.
Current docs cover MCP almost entirely from the consumer side. Connecting clients to mcp.apify.com, configuring tool selection, OAuth (sources/platform/integrations/ai/mcp/). The publisher side, how to wrap an MCP server as an Actor, monetize it, register it with the Apify MCP server, is a one-liner in
develop_ai_agents.mdand conceptual prose in the Academy actorization playbook. No step-by-step guide exists.Scope
A new how-to guide covering:
Companion work
Acceptance criteria
References