token-ninja routes deterministic shell commands locally — zero LLM calls, ~19µs latency. Works silently inside AI tools via MCP.
Setup
Choose your client, copy the config, then fill in any required environment variables.
claude mcp add token-ninja -- npx token-ninja mcp~/.codex/config.toml
[mcp_servers.token-ninja]
command = "npx"
args = ["token-ninja", "mcp"]
# codex mcp add token-ninja -- npx token-ninja mcp~/.cursor/mcp.json
{
"mcpServers": {
"token-ninja": {
"command": "npx",
"args": [
"token-ninja",
"mcp"
]
}
}
}.vscode/mcp.json
{
"servers": {
"token-ninja": {
"type": "stdio",
"command": "npx",
"args": [
"token-ninja",
"mcp"
]
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"token-ninja": {
"command": "npx",
"args": [
"token-ninja",
"mcp"
]
}
}
}claude_desktop_config.json
{
"mcpServers": {
"token-ninja": {
"command": "npx",
"args": [
"token-ninja",
"mcp"
]
}
}
}.trae/mcp.json
{
"mcpServers": {
"token-ninja": {
"command": "npx",
"args": [
"token-ninja",
"mcp"
]
}
}
}