What Are MCP Servers on Apify?
MCP (Model Context Protocol) servers are Apify actors that run in standby mode and expose tools via an HTTP endpoint for AI assistants like Claude Desktop, Cursor, and Windsurf. They connect large language models to real-world data sources -- APIs, databases, web scrapers, and intelligence feeds -- so AI agents can take actions beyond text generation.
MCP (Model Context Protocol) servers on Apify are actors that run continuously in standby mode and expose tools via an HTTP POST endpoint at /mcp. When you connect an AI assistant like Claude Desktop, Cursor, or Windsurf to an MCP server, the assistant gains the ability to call real-world data tools -- search APIs, web scrapers, database lookups, compliance checks -- as part of its reasoning process. Instead of just generating text, the AI can fetch live data, run analyses, and return structured results. ApifyForge publishes 93 MCP servers covering intelligence, compliance, market research, and developer tooling.
How MCP works
The Model Context Protocol is a JSON-RPC 2.0 based standard that defines how AI clients (like Claude Desktop) communicate with tool servers. The flow is:
- Discovery. The client sends an
initializerequest to the server's/mcpendpoint. The server responds with a list of available tools, each with a name, description, and input schema.
- Tool call. When the AI decides it needs data, it sends a
tools/callrequest with the tool name and parameters. The server executes the tool (calling APIs, running scrapers, querying databases) and returns structured results.
- Response. The AI incorporates the tool results into its reasoning and generates a response for the user.
| MCP Concept | What It Is | Example |
|---|---|---|
| Server | An always-on actor exposing tools | apifyforge/multi-source-intelligence-mcp |
| Tool | A callable function on the server | search_sec_filings, analyze_brand_sentiment |
| Transport | How client and server communicate | HTTP POST (standard), SSE (streaming) |
| Client | The AI application calling tools | Claude Desktop, Cursor, Windsurf, custom apps |
How to connect an MCP server
To connect an MCP server to Claude Desktop, add it to your Claude Desktop configuration file. The config file location depends on your operating system:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"apifyforge-intelligence": {
"url": "https://actors-mcp-server.apify.actor/mcp",
"headers": {
"Authorization": "Bearer YOUR_APIFY_API_TOKEN"
}
}
}
}For Cursor and Windsurf, add the server URL in the MCP settings panel. The URL format is always https://ACTOR_NAME.apify.actor/mcp and authentication uses a Bearer token with your Apify API token.
ApifyForge's 93 MCP servers
ApifyForge MCP servers are organized into functional categories, each providing multi-source intelligence by combining data from several APIs and data sources into a single tool call:
| Category | Servers | What They Do |
|---|---|---|
| Financial Intelligence | 12 | SEC filings, sanctions screening, AML checks, trade data |
| Market Research | 15 | Competitor analysis, pricing intelligence, market gaps |
| Compliance & Risk | 18 | GDPR scanning, export controls, regulatory tracking |
| Developer Tools | 10 | Schema validation, regression testing, cost optimization |
| Brand & Reputation | 8 | Review analysis, social listening, brand protection |
| Geopolitical & Climate | 14 | Supply chain risk, climate impact, geopolitical stability |
| Healthcare & Science | 9 | Drug pipeline, clinical trials, PubMed research |
| Other Specialized | 7 | Aviation safety, nuclear risk, food safety, water systems |
Each MCP server typically exposes 3-8 tools. For example, the Multi-Source Intelligence MCP exposes tools for SEC filing search, sanctions screening, company research, and competitive analysis -- all in a single server connection.
Pricing
MCP servers use the same Pay-Per-Event pricing as regular actors, but the "event" is a tool call. Typical pricing:
| Tool Complexity | Price Per Call | Example |
|---|---|---|
| Simple lookup | $0.05 | Sanctions name check, patent search |
| Multi-source query | $0.10 - $0.15 | Company deep research, competitor scan |
| Complex analysis | $0.20 - $0.25 | Portfolio risk assessment, supply chain audit |
You also pay Apify's standby compute cost, which is approximately $0.40/GB-hour. Most MCP servers run with 256MB-1GB of memory, so standby costs are $0.10-$0.40 per hour while the server is active.
MCP servers vs regular actors
Both MCP servers and regular actors run on Apify, but they serve different use cases:
| Feature | Regular Actor | MCP Server |
|---|---|---|
| Execution mode | On-demand (start, run, stop) | Always-on (standby mode) |
| Interface | REST API with JSON input/output | JSON-RPC 2.0 via /mcp endpoint |
| Primary consumer | Scripts, integrations, schedules | AI assistants (Claude, Cursor, Windsurf) |
| Latency | 5-30 seconds (container startup) | 100-500ms (already running) |
| Pricing | Per event (per result row) | Per event (per tool call) |
| Discovery | Apify Store listing | MCP initialize handshake |
| Multi-tool | One function per actor | 3-8 tools per server |
Use a regular actor when you need batch processing, scheduled runs, or integration with non-AI systems. Use an MCP server when you want an AI assistant to call tools interactively during a conversation.
Frequently asked questions
Do I need a paid Apify account to use MCP servers?
The free tier includes $5/month in platform credits, which is enough for light MCP usage (roughly 50-100 tool calls depending on the server). For regular use, a paid plan is recommended.
Can I build my own MCP server on Apify?
Yes. Any Apify actor can become an MCP server by enabling standby mode in .actor/actor.json and implementing the JSON-RPC 2.0 MCP protocol. The ApifyForge MCP Debugger at /tools/mcp-debugger helps you test your implementation.
What happens if the MCP server is down?
If a standby actor crashes, Apify automatically restarts it. During the restart (typically 3-10 seconds), MCP clients receive a connection error and retry. Claude Desktop retries automatically; Cursor shows a "tool unavailable" message.
Can I use multiple MCP servers at once?
Yes. Claude Desktop, Cursor, and Windsurf all support multiple MCP server connections simultaneously. Each server appears as a separate set of tools. You can connect financial intelligence, compliance, and market research servers at the same time.
How do MCP servers handle authentication?
MCP servers on Apify use Bearer token authentication. Your Apify API token is sent in the Authorization header with every request. The server validates the token before executing any tool call. You should never share your API token publicly.
Are MCP tool calls logged?
Yes. Every tool call is logged as an Apify actor run event. You can view the history in the Apify Console under the actor's Runs tab. ApifyForge Monitor at /monitor aggregates these across your fleet for centralized tracking.
Related guides
Getting Started with Apify Actors
To build an Apify actor, install Node.js 18+ and the Apify CLI, scaffold a project with apify create, write your logic inside Actor.main(), define an input_schema.json, and deploy with apify push. This guide walks through every step from zero to a published Apify Store listing.
Apify PPE Pricing Explained: Pay Per Event Model, Strategy, and Code Examples
Pay Per Event (PPE) is Apify's usage-based monetization model for actors on the Apify Store. Developers set a price per event (typically $0.001 to $0.50), call Actor.addChargeForEvent() in their code, and keep 80% of revenue while Apify takes 20%. This ApifyForge guide covers the 80/20 revenue split, actor.json configuration, charging code patterns, the 14-day price change rule, and pricing strategy by actor type.
How to Monetize Your Actors
To monetize Apify actors, start with Pay Per Event pricing at $0.01-$0.25 per result, then layer on tiered pricing for power users, free-tier funnels to drive adoption, and MCP server bundles that combine multiple actors into a single subscription. ApifyForge analytics tracks revenue per actor so you know which strategies work. This guide covers each revenue model with real pricing examples.
Actor Testing Best Practices
To test an Apify actor, define input/output test cases in a JSON fixture, run them with the ApifyForge test runner before every deploy, and set assertions on output shape, field counts, and error rates. The regression suite catches breaking changes by comparing current output against a saved baseline. This guide covers the full testing workflow from local validation to CI/CD integration.
Store SEO Optimization
Apify Store search ranks actors by title match, README keyword density, category tags, run volume, and a quality score out of 100. To rank higher, write a README that opens with a plain-language description of what the actor does, include target keywords in the first 100 words, set accurate categories in actor.json, and maintain a success rate above 95%. This guide breaks down every ranking factor and shows how ApifyForge tracks your score.
Managing Multiple Actors
To manage 10, 50, or 200+ Apify actors, use the ApifyForge fleet dashboard to monitor health, revenue, and quality scores across your entire portfolio in one view. Group actors by category, run bulk updates on pricing and metadata, set up failure alerts, and track maintenance pulse to catch stale actors before users complain. This guide covers fleet management workflows at every scale.
Cost Planning Tools: Calculator, Plan Advisor & Proxy Analyzer
How to use ApifyForge's cost planning tools to estimate actor run costs, choose the right Apify subscription plan, and pick the most cost-effective proxy type for each scraper.
AI Agent Tools: MCP Debugger, Pipeline Builder & LLM Optimizer
How to use ApifyForge's AI agent tools to debug MCP server connections, design multi-actor pipelines, optimize actor output for LLM token efficiency, and generate integration templates.
Schema Tools: Diff, Registry & Input Tester
How to use ApifyForge's schema tools to compare actor output schemas, browse the field registry, and test actor inputs before running — preventing wasted credits and broken pipelines.
Compliance Scanner, Actor Recommender & Comparisons
How to use ApifyForge's compliance risk scanner to assess legal exposure, the actor recommender to find the best tool for your task, and head-to-head comparisons to evaluate competing actors.
The ApifyForge Testing Suite
Five cloud-powered testing tools for Apify actors: Schema Validator, Test Runner, Cloud Staging, Regression Suite, and MCP Debugger. How they work together and when to use each one.
The Complete ApifyForge Tool Suite
All 14 developer tools in one guide: testing, schema analysis, cost planning, compliance scanning, LLM optimization, and pipeline building. What each tool does, when to use it, and how they work together.
What Is an Apify Actor?
An Apify actor is a serverless cloud program that runs on the Apify platform. It accepts JSON input, executes a task (scraping, data processing, API calls, or AI tool serving), and produces structured output in datasets, key-value stores, or request queues. Actors are packaged as Docker containers and can be run via API, scheduled, or chained together.
How to Choose the Right Apify Actor
With over 3,000 actors on the Apify Store, choosing the right one for your task requires evaluating success rates, run history, pricing, maintenance frequency, and input schema quality. This guide provides a decision framework for selecting actors based on measurable quality metrics, plus tools to automate the comparison process.
How to Manage a Large Apify Actor Portfolio
Managing 10 Apify actors is straightforward. Managing 50 requires dashboards and cost tracking. Managing 200+ demands automated regression testing, schema validation, revenue analytics, and failure alerting. This guide covers the tools, processes, and hard-won lessons from scaling an actor portfolio from a handful to over 320 actors.