Kosha Model Discovery runs on your own machine — the client starts it, so there is no endpoint to ping. 669 installs a week from npm. Last commit 19 Sep 2026.
Tells your agent which model to use and what it costs, across 45 providers and local runtimes.
Today is the operative word: we check Kosha Model Discovery every 15 minutes and re-read its code on every release. Watch it and you find out the day that stops being true.
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add kosha-discovery -- npx -y @sriinnu/kosha-discovery
{
"mcpServers": {
"kosha-discovery": {
"args": [
"-y",
"@sriinnu/kosha-discovery"
],
"command": "npx"
}
}
}
[mcp_servers.kosha-discovery]
command = "npx"
args = ["-y", "@sriinnu/kosha-discovery"]
{
"mcpServers": {
"kosha-discovery": {
"args": [
"-y",
"@sriinnu/kosha-discovery"
],
"command": "npx"
}
}
}
{
"mcpServers": {
"kosha-discovery": {
"args": [
"-y",
"@sriinnu/kosha-discovery"
],
"command": "npx"
}
}
}
This one needs environment variables set before it will start:
ANTHROPIC_API_KEY (Anthropic API key. Optional — without it, Anthropic models are read from the public models.dev / LiteLLM catalogs instead of the live API.), OPENAI_API_KEY (OpenAI API key. Optional — falls back to the public catalogs when absent.), GOOGLE_API_KEY (Google AI Studio / Gemini API key. Optional.), OPENROUTER_API_KEY (OpenRouter key. Optional for discovery (unauthenticated requests are rate-limited), required to execute through OpenRouter.), XAI_API_KEY (xAI (Grok) API key. Optional.), TYPESAFE_API_KEY (TypeSafe System One (Jev) API key. Optional.), KOSHA_CACHE_TTL_HOURS (Override the discovery cache TTL. Defaults to 24 hours.).
The author declared them in the registry entry; get the values from the project itself.
Local profiler: which usage window took you out, and where your agent's money goes
Track AI agent costs, detect waste, optimize models, and prove ROI. 23 MCP tools across 10 domains.
Fast local and federated OSS code search and repo discovery for AI agents
An MCP server that provides read access to your cloud storage providers, bank accounts and more.
AI tools that run on your own machine: what each does, installs like, and costs.
Cost-parity bench for agent loops: $/success and time across model mixes, with verifiable receipts.
Local runtime provenance for coding agents. Inspect sessions and explain which one owns a process.
Tells you how you and your Claude agent actually work together. Local-only, no API keys.
Answers built from our own checks of this server.