To let Claude or ChatGPT retrieve product data through Apify, connect the AI client to Apify’s hosted MCP server at https://mcp.apify.com, authorize it, then give it access to an Actor suited to your task. The setup differs by client: ChatGPT and Claude Desktop use remote connectors, while Claude Code has a separate CLI/plugin route. MCP provides access to Apify tools; it does not guarantee that an Actor’s results are current, complete, or accurate. Those depend on the Actor, its inputs, the source site, and when its run executes.
What Apify’s MCP server does—and what it does not
Apify’s hosted MCP server connects compatible AI clients to Apify platform capabilities, including discovering and running Actors, accessing storage and results, and searching Apify documentation. An Actor is the component that performs a specific data task, such as collecting product listings from a supported source. The MCP connection makes those capabilities available to the client; it does not itself collect product data or promise a freshness interval.
As an Amazon Associate I earn from qualifying purchases.
Apify describes its ChatGPT integration as a way to connect ChatGPT to its Actor library through MCP. Whether a result reflects current product information depends on the Actor’s behavior, its input and access to the source, and the timing and outcome of the run. Check the returned data and the Actor’s details before relying on it.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Apify’s current documentation lists a hosted-server limit of up to 30 MCP requests per second per user. That is an operational limit, not a guarantee of response speed. Apify’s ChatGPT integration page labels the integration beta and notes that it may be slow and that Developer mode availability can vary. These details were documented as of October 2026 and may change.
#1 Best Overall
Choose the connection route for your client
| Route | Best fit | Authentication and setup | Local runtime |
|---|---|---|---|
| Hosted remote MCP | ChatGPT and Claude Desktop, when their connector flows support remote MCP | Apify recommends browser-based OAuth; bearer-token configuration is also documented | Not required |
| Claude Code | Claude Code CLI users | Use Apify’s Claude Code plugin or the documented CLI installation route | Uses the CLI environment |
| Local stdio MCP | Clients that cannot connect to remote MCP servers | Run the local server with an Apify API token | Required; documented local setup requires Node.js v18 or higher |
The hosted service supports inferred output schemas for Actor results; Apify says local stdio does not. Consult Apify’s current instructions before installing or updating a CLI or package because client support and commands can change.
Connect Apify to ChatGPT
Apify’s documented ChatGPT route uses a custom connector and OAuth. The exact labels or availability can vary with ChatGPT’s current interface and account access.
- In ChatGPT, open Settings and enable Developer mode, if available for your account.
- Go to Settings > Apps & Connectors and choose Create.
- Enter a connector name and description. For the server URL, use
https://mcp.apify.comfor the standard tool set, or a selected-tools URL if you want to limit which tools are exposed. - Choose OAuth as the authentication method and complete the Apify authorization in the browser.
- Start a new chat, open the plus button, choose More, and select the Apify connector. ChatGPT may request permission the first time you use individual tools.
Apify’s guide says a connector’s selected tools cannot be changed after creation. If you want a different tool set, create another connector with the revised URL rather than expecting to edit the existing one.
Recommended Free Tools
Add Apify as a connector in Claude Desktop
Claude Desktop connects to Apify’s hosted server through its custom connector flow, not by placing a remote server URL in a local configuration file.
- Open Claude Desktop’s settings and find the area for connectors or integrations.
- Add a custom connector and enter
https://mcp.apify.com. If an Apify listing appears in the connector directory, that is another documented route. - Complete the browser-based authorization with Apify when prompted.
- Open a conversation and enable the Apify connector or its tools when Claude offers them.
Claude’s interface can change, so use the connector controls currently shown in your installation. Do not apply Claude Code CLI instructions to Claude Desktop.
Use Apify with Claude Code
Claude Code is a separate command-line client. Apify documents a Claude Code plugin route and lists claude-code among the clients supported by its CLI installation command. Follow the current Apify Claude integration and CLI instructions for the plugin or installation flow; the Desktop connector steps above are not a substitute.
Rank #3
Apify’s setup references are Claude integration overview and the Apify CLI reference. Check the current command and authentication prompts there before using them, rather than copying a command from an older setup.
Use local stdio when remote MCP is unavailable
If your client cannot connect to remote MCP servers, Apify documents a local stdio server. It runs on your machine and uses an Apify API token. The documented package command is:
npx -y @apify/actors-mcp-serverConfigure the client to launch that command and provide the token as the APIFY_TOKEN environment variable, using the client’s supported environment-variable configuration. Apify specifies Node.js v18 or higher for this local environment. Avoid placing a token directly in a configuration file that may be shared, committed to version control, or exposed to other users; use a protected environment or the client’s secure secret mechanism where available.
Local stdio requires maintaining the local runtime and package, and it does not provide the inferred Actor-result output schemas available through Apify’s hosted service. For the hosted connection, Apify recommends remote OAuth; it also documents bearer-token configuration. Use the authentication method supported by your client and avoid granting or storing more access than necessary.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Limit available tools and control Actor runs
Expose only the tools you need
Apify supports narrowing the MCP tool set with a tools query parameter in the server URL. Its CLI also offers a --tools option. For ChatGPT, choose the intended tool set when creating the connector, since Apify says that selection cannot later be edited. Limiting tools can make the client’s available actions more focused, but it does not replace limits on an Actor run.
Set run limits in call options
When asking the MCP server to run an Actor, set execution controls in callOptions. Apify’s agent onboarding guide identifies controls such as maxTotalChargeUsd, maxItems, timeout, and memory. Do not assume fields with similar names inside an Actor’s input impose those limits: the execution controls belong in callOptions.
Choose limits appropriate to the task and inspect the Actor’s input requirements before running it. A product lookup that needs a small sample may not need the same item or time allowance as a broad collection. These controls constrain the run; they do not guarantee a particular result or freshness.
Authentication, access, and privacy considerations
Apify documents several tools that do not require authentication: search-actors, fetch-actor-details, search-apify-docs, and fetch-apify-docs. Running Actors and accessing storage or run data require authentication. Apify also says its MCP server excludes full-permission Actors and rental Actors.
Apify’s telemetry documentation says basic tool-call and client attributes are collected by default and documents an opt-out URL parameter, telemetry-enabled=false. This describes Apify’s stated telemetry behavior; it is not an independent privacy audit. Review Apify’s current MCP documentation and your organization’s data policies before connecting sensitive workflows.
Free tools Windows power users keep installed
One-click scans. No signup required.
Make product-data requests specific
After connecting, name the kind of data you need and the source or Actor you want used. For example, ask the client to find an Actor suited to collecting product listings from a particular site, provide the relevant search terms or category, run it with a modest item and time limit, and report when the data was retrieved. If you already know the Actor, specify it and supply its required input.
Review the Actor’s details and output rather than treating the AI’s summary as a verified catalog. Product availability, price, variants, and other fields can differ by source and collection time. MCP makes the tools available; it cannot make a source expose information that is absent or inaccessible.
Quick Recap
Official setup references
- Apify MCP server — hosted and local setup, tool selection, authentication, limits, and telemetry.
- Apify ChatGPT integration — ChatGPT connector setup and current beta notes.
- Claude integration overview — Claude Desktop and Claude Code paths.
- Apify CLI reference — CLI commands and client installation support.
- Apify for AI agents — tool selection and Actor run controls.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
