Connect Cursor, VS Code with GitHub Copilot, or Gemini CLI
These clients can connect directly to Memini over Streamable HTTP.
Start Memini, then obtain a token with read scope for the
workspace you want the assistant to search. MCP
setup and token creation
Prepare the client environment
Load the workspace token into the environment that launches the
client. The examples assume the token is saved in
~/.config/memini/mcp.token. Substitute your token file.
Bash or Zsh
export MEMINI_READ_TOKEN="$(cat "$HOME/.config/memini/mcp.token")"PowerShell
$env:MEMINI_READ_TOKEN = (Get-Content -Raw (Join-Path $HOME '.config/memini/mcp.token')).Trim()Quit any running editor, then launch cursor .,
code ., or gemini from that shell so the
client inherits the variable. Merge the memini entry below
into your existing settings to preserve other servers.
The examples use Memini's default local URL. In a remote workspace or
dev container, 127.0.0.1 refers to the environment making
the connection. Supply the token there and use your operator's HTTPS
endpoint or a local tunnel to Memini. To check network reachability
without sending a token, open http://127.0.0.1:6364/readyz
from the same environment, adjusting the address as needed. A ready
server returns HTTP 200.
Cursor
Add this entry to your personal ~/.cursor/mcp.json or
the project's .cursor/mcp.json. Cursor resolves
${env:MEMINI_READ_TOKEN} in HTTP headers. Cursor MCP configuration
{
"mcpServers": {
"memini": {
"url": "http://127.0.0.1:6364/mcp",
"headers": { "Authorization": "Bearer ${env:MEMINI_READ_TOKEN}" }
}
}
}Open Cursor's MCP settings, enable memini, then use
Agent chat to run the check below.
VS Code with GitHub Copilot
For a session using the Copilot harness, edit
~/.copilot/mcp-config.json (or
$COPILOT_HOME/mcp-config.json when set). The workspace
equivalent is .mcp.json at the project root. This format
supports ${MEMINI_READ_TOKEN} inside the header value. The
older Local harness uses a separate editor-managed
.vscode/mcp.json format. VS
Code MCP configuration reference
{
"mcpServers": {
"memini": {
"type": "http",
"url": "http://127.0.0.1:6364/mcp",
"headers": { "Authorization": "Bearer ${MEMINI_READ_TOKEN}" },
"tools": ["*"]
}
}
}In Chat, select the Copilot harness. Run Chat: Open
Customizations from the Command Palette, select MCP
Servers and start or enable memini. Use
Show Output in the server menu to inspect connection
failures. VS
Code server controls
Gemini CLI
Add this entry to your personal ~/.gemini/settings.json
or the project's .gemini/settings.json. Gemini CLI selects
Streamable HTTP with httpUrl and expands environment
references when it loads the settings. Gemini CLI MCP
transports · Gemini CLI
settings
{
"mcpServers": {
"memini": {
"httpUrl": "http://127.0.0.1:6364/mcp",
"headers": { "Authorization": "Bearer ${MEMINI_READ_TOKEN}" }
}
}
}Run gemini mcp list from the configured shell and look
for memini marked connected. Start gemini and
run the check below.
Verify workspace access
Ask the assistant “Call the Memini status tool with no arguments. Show the workspace names from its result.” Check the tool call and confirm that your workspace appears. An empty workspace can have zero records. If the connection returns HTTP 401, check the client process received the token value and that the token is current. If your workspace is missing, check which workspace the token grants access to. Memini status tool