Offline MCP server that predicts LLM call cost and recommends the cheapest capable model before you call it. No API keys, no network. npx @workswarm/modelcostsaver
Setup
Choose your client, copy the config, then fill in any required environment variables.
claude mcp add modelcostsaver -- npx -y @workswarm/modelcostsaver~/.codex/config.toml
[mcp_servers.modelcostsaver]
command = "npx"
args = ["-y", "@workswarm/modelcostsaver"]
# codex mcp add modelcostsaver -- npx -y @workswarm/modelcostsaver~/.cursor/mcp.json
{
"mcpServers": {
"modelcostsaver": {
"command": "npx",
"args": [
"-y",
"@workswarm/modelcostsaver"
]
}
}
}.vscode/mcp.json
{
"servers": {
"modelcostsaver": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@workswarm/modelcostsaver"
]
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"modelcostsaver": {
"command": "npx",
"args": [
"-y",
"@workswarm/modelcostsaver"
]
}
}
}claude_desktop_config.json
{
"mcpServers": {
"modelcostsaver": {
"command": "npx",
"args": [
"-y",
"@workswarm/modelcostsaver"
]
}
}
}.trae/mcp.json
{
"mcpServers": {
"modelcostsaver": {
"command": "npx",
"args": [
"-y",
"@workswarm/modelcostsaver"
]
}
}
}