MCP Setup
Connect AI agents and IDEs to the Tapaya documentation and REST API over the Model Context Protocol.
The Tapaya Docs MCP server lets AI agents and IDEs search this documentation and explore the Tapaya REST API directly, without leaving your editor. It is read-only, public, and requires no API key or secret.
Server URL
The server speaks Streamable HTTP and works with any MCP client.
Connect your client
Install the plugin (recommended):
claude plugin marketplace add tapayadot/tapaya-mcp
claude plugin install tapaya-docs@tapayaOr add the server directly without the plugin:
claude mcp add --transport http tapaya-docs https://docs.tapaya.com/mcpThen check it is connected:
claude mcp listOr add the following to ~/.cursor/mcp.json manually:
{
"mcpServers": {
"tapaya-docs": {
"url": "https://docs.tapaya.com/mcp"
}
}
}Or add the following to .vscode/mcp.json in your workspace manually:
{
"servers": {
"tapaya-docs": {
"type": "http",
"url": "https://docs.tapaya.com/mcp"
}
}
}Add the following to your Claude Desktop MCP config:
{
"mcpServers": {
"tapaya-docs": {
"url": "https://docs.tapaya.com/mcp"
}
}
}MCP is an open protocol. Point any Streamable HTTP client at the server url https://docs.tapaya.com/mcp. You can list the available tools with a raw request:
curl -s https://docs.tapaya.com/mcp \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'Available tools
| Resource | Tool | Description |
|---|---|---|
| API tools | search_docs | Full-text search across all documentation pages. |
get_page | Fetch the full Markdown of a page by its url. | |
list_pages | List every page as title, url, and description. | |
search_api_endpoints | Search the Tapaya REST API by keyword. | |
get_api_endpoint | Get full parameters and responses for one endpoint. |
Notes and safety
This server is read-only. It never calls the live Tapaya API and never asks for API keys or secrets.
When you combine multiple MCP servers, exercise the usual caution around prompt injection from tool output, and enable human confirmation for tools where your client supports it.
Found a problem or want another tool? Open an issue on the docs repository.