MCP Laya runs on your own machine — the client starts it, so there is no endpoint to ping. 194 installs a week from pypi. Last commit 12 Oct 2026.
Safe-by-default MCP for Laya: fast, local, typed decisions (classify/score/yes-no), 100+ langs.
Today is the operative word: we check MCP Laya 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 mcp-laya -- uvx mcp-laya
{
"mcpServers": {
"mcp-laya": {
"args": [
"mcp-laya"
],
"command": "uvx"
}
}
}
[mcp_servers.mcp-laya]
command = "uvx"
args = ["mcp-laya"]
{
"mcpServers": {
"mcp-laya": {
"args": [
"mcp-laya"
],
"command": "uvx"
}
}
}
{
"mcpServers": {
"mcp-laya": {
"args": [
"mcp-laya"
],
"command": "uvx"
}
}
}
This one needs environment variables set before it will start:
LAYA_DEFAULT_MODEL (Checkpoint to use: auto (Router picks) | english | multilingual | typed-decisions.), LAYA_MODELS (Comma-separated allowlist of checkpoints that may load. Empty = all.), LAYA_ALLOW_DOWNLOAD (Allow the one-time Hugging Face checkpoint download. Off by default (offline).), LAYA_MIN_CONFIDENCE (Flag answers below this calibrated confidence (0 = never flag).), LAYA_DEVICE (Inference device: auto | cpu | cuda.).
The author declared them in the registry entry; get the values from the project itself.
Snap decisions for coding agents: local or hosted classify/check/score/rank via MCP.
Typed decisions (choice, score, yes/no) from Laya, Julia-1 and clef-flash on your own GPU
Local Laya and Von typed decisions over TypeSafe-compatible MCP.
Typed noul / choice / score judgments from a local Laya model, with token-budget preflight.
Local Phocinae decision-engine MCP stdio server: fail-closed approval gate + classify/route/score.
Typed decisions with Jev / System One via OpenRouter or TypeSafe: classify, verify, rerank, decide.
Typed Jev decisions in one batched tool: choice, score, and yes/no probabilities.
Local 2B and 0.8B decision models: calibrated yes/no, choice and score answers for your agent.
Answers built from our own checks of this server.