Discover MCP endpoints
Paste a website, direct MCP endpoint, or metadata JSON. The playground probes the usual discovery paths and connects to the first remote endpoint that works.
MCP Playground
Discover, connect, inspect primitives, and run MCP calls from one protocol workbench.
remote transport
Connect to an MCP server
Enter a site, endpoint, or metadata JSON. Discovery runs first, then the first working endpoint is connected automatically.
Transport
Popular servers
Optional shortcuts for hosted MCP servers.
Limit: 5 connections per 60 minutes.
Connection log
stdoutWaiting for Connect.
Model Context Protocol testing
A browser playground for remote MCP servers
The MCP Playground helps developers try hosted Model Context Protocol servers before connecting them to agents, IDEs, or internal tools. Paste a website, MCP endpoint, or server metadata. The playground runs discovery, connects to a Streamable HTTP or SSE endpoint, lists tools, resources, resource templates, and prompts, then opens a console for tool calls, resource reads, prompt fetching, OAuth flows, headers, notifications, response payloads, and JSON-RPC errors. It is for remote MCP validation, not local stdio sessions, so teams can check real endpoints, auth behavior, schemas, and event streams from one browser page.
Inspect server capabilities
Check server info, protocol version, transport mode, tools, resources, resource templates, prompts, initialized notifications, and streamed events.
Test MCP operations
Call tools with a generated form or raw JSON, read resources, get prompts, inspect responses, and copy the raw JSON-RPC payload when something fails.
Check auth behavior
Use OAuth, bearer tokens, and custom headers to test protected MCP servers while keeping connection state tied to the current playground session.
What the MCP Playground validates
Discovery: Tests common MCP discovery paths, direct endpoints, and pasted metadata so teams can see whether a server is findable from a known URL.
Protocol data: Lists tools, resources, resource templates, prompts, capabilities, protocol version, transport mode, initialized notifications, and streamed events.
Operations: Runs MCP tool calls with raw JSON or generated forms, reads resources, gets prompts, and records responses in a developer timeline.
Authentication: Supports OAuth authorization, bearer tokens, custom headers, and per-session connection state for protected MCP servers.
FAQ
MCP Playground questions
What is an MCP Playground?
An MCP Playground is a developer tool for connecting to Model Context Protocol servers, seeing which tools, resources, prompts, and notifications they expose, and trying JSON-RPC calls before wiring the server into an agent or app.
What MCP transports does this playground support?
The playground supports remote MCP servers over Streamable HTTP and SSE. It does not support local stdio servers because this page is meant for hosted endpoints that can be reached, authenticated, and tested from a browser session.
Can the MCP Playground test OAuth-protected MCP servers?
Yes. If an MCP endpoint asks for authorization, the playground can start the OAuth flow, return to the same session, and reconnect after consent. Developers can also add bearer tokens, custom headers, and OAuth parameters without saving secrets as permanent backend records.
Who should use the MCP Playground?
The MCP Playground is for developers, platform teams, and AI infrastructure teams that need to check MCP discovery, auth, tools, resources, prompts, schemas, responses, and protocol events.