Give AI agents
access to live APIs.
Discover APIs, inspect schemas, and call live data through one hosted MCP server.
(MCP-compatible)
MCP ServerHosted endpoint
with built-in tools
- Hosted MCP endpointOne endpoint to connect your agent
- 3 marketplace toolsSearch, inspect, and call
- Server-side credentialsUse directly in compatible clients
Watch a question become an answer backed by facts.
On the left, an assistant asks for supplier information. Follow how ApiClay finds an API, checks its fields, and retrieves the facts used in the answer on the right.
From your input to a useful result.
- Choose a sample input
- Apply the feature’s workflow
- Inspect the result
This example runs entirely in your browser with illustrative data. Use the real feature’s guide and listing to check supported inputs, access rules, and pricing.
Explore the example, then use the setup guide to try the real feature.
Project setup and what to check
Input: A supplier question, an authenticated MCP client, and a buyer API key. Choose a listing whose access rules and data fit your research.
Workflow: Search for a catalog API, inspect its required inputs and pricing, then approve a call in your client. Keep the returned source data alongside the agent’s summary.
Result: A source-backed comparison your application can review. The agent works with the published API contract while upstream credentials stay on the server.
Confirm the returned fields answer your question, handle missing records, and review the execution charge before extending the workflow.
From a question to a live answer.
Go from a natural language request to real data in three simple steps.
- 1
Connect your client
Add your ApiClay API key to your MCP-compatible client.
- 2
Discover the right API
Use the search tools to find the best API for your task.
- 3
Review the response
Inspect the schema and call the API to get live data back to your agent.
A clear contract for every tool call.
Connect to a single, hosted MCP endpoint using your API key.
{
"mcpServers": {
"apiclay": {
"url": "https://api.apiclay.com/v1/mcp",
"headers": {"Authorization": "Bearer YOUR_API_KEY"}
}
}
}1. apiclay_search_apis
Find a supplier catalog API.
2. apiclay_describe_api
Inspect its inputs, schema, and pricing.
3. apiclay_call_api
Call the selected route with valid inputs.Upstream credentials stay server-side
Connect to third-party APIs through ApiClay. Provider credentials stay on the server, not in your client or prompts.
Keep control of tool execution
Your MCP client controls tool approvals. Review consequential actions before execution.
A Model Context Protocol server for real data
An AI agent needs current information and a clear tool contract to answer useful questions. ApiClay's hosted MCP server connects that agent to marketplace listings backed by databases, spreadsheets, HTTP services, and live websites. Each listing supplies the documented inputs and outputs the agent needs to choose and call a route.
Use the same buyer API key for REST calls and MCP execution. ApiClay handles upstream credentials on the server, so your agent can work through the published contract without receiving the publisher's database password, API token, or proxy credentials. Your MCP client still controls which tool actions the agent may perform.
Search the marketplace through tools
apiclay_search_apis finds candidate APIs and apiclay_describe_api exposes their contracts and pricing.
How it works
These discovery tools are free. apiclay_call_api executes the chosen route, applying its authentication, quotas, and billing rules. The separation lets an agent inspect a listing before spending credits on a call.
Reach browser-rendered APIs through one contract
Published routes that enable JavaScript or automatic rendering work transparently through apiclay_call_api.
How it works
When an approved route needs scripted clicks, form entry, session bootstrap, or CAPTCHA handling, the gateway executes those configured steps and returns clean JSON or Markdown to your agent.
Use documentation that coding assistants can read
Public listings provide Markdown documentation through /docs/<vendor>.md and listing-level llms.txt mirrors.
How it works
Supply these references to a coding assistant when building an integration. They complement the interactive Reference and Pricing tabs on each API playground rather than requiring a separate documentation portal.
Ground answers in current source responses
Retrieve approved database records or request clean Markdown from a web source before generating an answer.
How it works
Structured schemas and smaller outputs help an agent identify relevant facts. Source access improves grounding, while your application remains responsible for checking evidence, handling missing data, and approving consequential actions.
Where this fits into your work.
Supplier research
Find supplier APIs, inspect their fields, and gather current catalog information for a procurement comparison. Call browser-rendered routes when the permitted source loads its catalog dynamically.
Support copilots
Read an approved entitlement API and relevant help content before drafting a response. Keep database credentials outside the model's context.
AI coding workflows
Give a coding assistant a listing's Markdown specification and use MCP to inspect the data contract while developing a REST integration.
Start with a focused test.
Use these steps to set up MCP Server and review the result before making it part of a larger workflow.
Create an API key →- 1
Create a buyer API key
Sign in and generate an apiclay_* key in API Keys. Keep it in your MCP client's protected credential settings and fund the wallet before paid execution.
- 2
Connect your MCP client
Choose a client that supports authenticated remote Streamable HTTP. Add the gateway's /v1/mcp endpoint and send your key as an Authorization: Bearer credential.
- 3
Discover, inspect, then call
Ask the agent to find a suitable API, inspect its schema and pricing, and call the selected route. Review the result before using it in a wider workflow.
Understand each workflow’s cost.
Match the feature’s limits and billing model to the work you want to run.
Explore API listings →Marketplace search and schema discovery are free. API execution follows the selected listing's trial, plan, and usage rules; browser rendering, verified CAPTCHA solving, and proxy use can add charges when applicable.
Free discovery lets an agent check required inputs, response schemas, and route pricing before making a paid call. Review the live listing and wallet ledger for applicable per-request and add-on fees.
Open an API listing for its Reference, Playground, and Pricing. Check the selected route’s allowances and applicable infrastructure charges before increasing usage.
Frequently asked questions
Quick answers about the ApiClay MCP Server.
Do I need to run a local MCP server?
No. ApiClay hosts the server. Use an MCP client that supports the remote endpoint and bearer authentication; a client with only local-server support may need its own supported bridge.
Can I connect Claude or Cursor?
Use a version and configuration that supports authenticated remote MCP servers. Add ApiClay's endpoint and bearer key using that client's connection settings. Available tools and approval controls are managed by the client.
Does the agent receive publishers' private API keys?
Publishers' upstream credentials stay server-side. The agent authenticates to ApiClay with your buyer key and receives the published route contract and response. Protect your buyer key as an execution credential.
How is MCP different from calling the REST API?
REST is suited to explicit application requests. MCP presents discovery and execution as tools an AI client can use. Both use the same gateway-backed marketplace and buyer wallet.
Connect your next agent to real data.
Create an API key and start using the ApiClay MCP Server today.