Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.mcpjungle.com/llms.txt

Use this file to discover all available pages before exploring further.

This page shows practical registration examples for commonly used MCP servers. It complements the transport-specific guides: Use direct CLI flags when a server only needs a name and URL. Use JSON config files when the server needs advanced configurations like authentication, OAuth, custom headers, environment variables, or stateful sessions. For any JSON example on this page, register it with: 
mcpjungle register -c ./<filename>.json

​
Hosted HTTP servers

​
Context7: fastest possible registration

If the upstream server is public and does not require extra configuration, register it directly with CLI flags: 
mcpjungle register --name context7 --url https://mcp.context7.com/mcp
This is the simplest pattern for hosted Streamable HTTP servers.

​
DeepWiki: simple config file

If you prefer to keep server registrations in version control, use a JSON config file even for simple HTTP servers: 
{
  "name": "deepwiki",
  "transport": "streamable_http",
  "url": "https://mcp.deepwiki.com/mcp",
  "description": "DeepWiki MCP server"
}

​
GitHub: bearer token from an environment variable

For servers that require a static token, it is recommended to keep the secret out of the file and inject it from the shell environment: 
{
  "name": "github",
  "transport": "streamable_http",
  "url": "https://api.githubcopilot.com/mcp",
  "description": "GitHub MCP server",
  "bearer_token": "${GITHUB_PAT}"
}

export GITHUB_PAT=your_token_here
This is the pattern to reuse for hosted servers that accept standard bearer authentication.

​
Figma: upstream OAuth during registration

Some hosted servers use OAuth instead of static API tokens. Figma is a good example: 
{
  "name": "figma",
  "transport": "streamable_http",
  "url": "https://mcp.figma.com/mcp",
  "description": "Figma MCP server"
}
If the upstream server advertises OAuth, the CLI opens a browser, you approve access, and Mcpjungle completes registration automatically.

​
Todoist: OAuth flow from a direct CLI command

Todoist is a good example of a hosted server where the fastest path is still a direct CLI command: 
mcpjungle register --name todoist --url https://ai.todoist.net/mcp
Todoist enforces OAuth. Running this command starts the upstream OAuth flow in your browser, and Mcpjungle completes the registration after you approve access.

​
Local STDIO servers

​
Filesystem: minimal STDIO example

The official filesystem server is the canonical example of local STDIO registration: 
{
  "name": "filesystem",
  "transport": "stdio",
  "command": "npx",
  "args": ["-y", "@modelcontextprotocol/server-filesystem", "."],
  "description": "Filesystem MCP server"
}
This pattern is useful for local tools that do not need extra secrets or long-lived sessions. If you run Mcpjungle inside Docker and want to use the filesystem server, see the Docker caveat in Add STDIO MCP servers.

​
Playwright: STDIO server with stateful sessions

Playwright requires persistent sessions so each tool call can reuse the same running browser process: 
{
  "name": "playwright",
  "transport": "stdio",
  "command": "npx",
  "args": ["-y", "@playwright/mcp@latest"],
  "description": "Playwright MCP server",
  "session_mode": "stateful"
}
You must use session_mode: "stateful" for the Playwright MCP server. In stateless mode, Playwright does not work properly because each tool call starts a fresh process instead of reusing the active browser session.

​
n8n: STDIO server with required environment variables

Some STDIO servers need environment variables at process startup: 
{
  "name": "n8n",
  "transport": "stdio",
  "command": "npx",
  "args": ["-y", "n8n-mcp"],
  "description": "n8n MCP server",
  "env": {
    "MCP_MODE": "stdio",
    "LOG_LEVEL": "error"
  }
}
Add secrets the same way: 
{
  "env": {
    "API_KEY": "${N8N_API_KEY}"
  }
}

​
Time: small uvx-based server

If your STDIO server is Python-based, uvx is often the cleanest launcher: 
{
  "name": "time",
  "transport": "stdio",
  "command": "uvx",
  "args": ["mcp-server-time"],
  "description": "Time MCP server",
  "session_mode": "stateful"
}
This is a good template for small Python-distributed MCP servers.

​
SSE transport

SSE is a deprecated MCP transport and is discouraged in Mcpjungle. Prefer Streamable HTTP for remote servers unless you specifically need to connect to an older SSE-based deployment.
If you still need to register an SSE server, use a JSON config file: 
{
  "name": "test-mcp",
  "transport": "sse",
  "url": "http://host.docker.internal:9000/sse",
  "description": "Legacy SSE MCP server"
}
This is mainly useful for compatibility with older MCP servers that have not migrated to Streamable HTTP yet.

HTTP servers

Register hosted MCP servers over Streamable HTTP.

STDIO servers

Register local MCP servers that run as child processes.

Session modes

Decide when to use stateless versus stateful server sessions.

Config file reference

See the exact JSON config formats and environment substitution rules.