MCP server that shrinks logs, JSON, docs and web pages before your AI reads them. Offline, deterministic, ~3ms, no API key. 66% fewer tokens on a mixed dev corpus, 90%+ on logs.
Setup
Choose your client, copy the config, then fill in any required environment variables.
claude mcp add latentgate -- uvx latent-gate~/.codex/config.toml
[mcp_servers.latentgate]
command = "uvx"
args = ["latent-gate"]
# codex mcp add latentgate -- uvx latent-gate~/.cursor/mcp.json
{
"mcpServers": {
"latentgate": {
"command": "uvx",
"args": [
"latent-gate"
]
}
}
}.vscode/mcp.json
{
"servers": {
"latentgate": {
"type": "stdio",
"command": "uvx",
"args": [
"latent-gate"
]
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"latentgate": {
"command": "uvx",
"args": [
"latent-gate"
]
}
}
}claude_desktop_config.json
{
"mcpServers": {
"latentgate": {
"command": "uvx",
"args": [
"latent-gate"
]
}
}
}.trae/mcp.json
{
"mcpServers": {
"latentgate": {
"command": "uvx",
"args": [
"latent-gate"
]
}
}
}