Configure
Each MCP server in your Hub can be customised through the server detail page. Open MCP Servers, click on a server name, and use the options described below.
Environment variables
Key-value pairs injected into the MCP server process at startup. Use this for API keys, tokens, paths, or any other configuration that the server reads from the environment.
Values are stored encrypted. After saving, a value is displayed as a masked placeholder unless you click to reveal it.
Command and arguments (stdio only)
For stdio servers, you can override the default command and argument list. Each argument goes on a separate line. Changes take effect on the next MCP Proxy session.
Custom instructions
A free-text field that is prepended to the system prompt for every tool call made through this server. Use it to give the model context about the server's purpose, preferred output format, or constraints.
Example:
Always respond in the same language as the user's query.
Only use the read_file tool; never use write_file.
Custom instructions are applied by the MCP Proxy before forwarding requests and are not visible to the upstream MCP server.
Credential files (Files tab)
Some MCP servers (for example, Google Analytics MCP) require a file on the filesystem — a service-account JSON, certificate, or other config file — referenced via an environment variable such as GOOGLE_APPLICATION_CREDENTIALS.
The Files tab on the server detail page lets you handle this:
- Upload a file — upload the credential or config file (up to 100 KB). Uploaded files are encrypted at rest and scoped to your Hub. Up to 5 files per Hub.
- Add a file mapping — link a file to an environment variable name for this server (e.g.,
GOOGLE_APPLICATION_CREDENTIALS). At runtime, the MCP Proxy writes the file to the server's sandbox and sets the env var to point to it.
One uploaded file can be reused across multiple MCP servers.