dockndevai/mcp-laya

A **safe-by-default** Model Context Protocol server for **Laya** — a fast, non-autoregressive **System-1 decision engine**. It gives an agent a *thinking* primitive: typed decisions — **classify, score, yes/no** — over any state (text, an email, a ticket, a JSON object), in a **single local forward pass (~33 ms)**, across **100+ languages**, with **no text generation — nothing to parse and nothing to hallucinate**, and a **calibrated confidence** on every answer.
Instead of burning a slow, costly LLM round-trip on "which team should handle this? is it urgent? is this a refund request?", the agent calls a typed tool that answers **locally, in milliseconds, offline**. It's the safest server in the suite — laya only *reads* a state and returns a decision; it changes nothing.
Part of the dockndevai MCP server suite — one governance model across all of them. (This is the first Python server in the suite; the rest are Node/TS.)
What it gives an agent
| Tool | For |
|---|---|
| decide | answer several typed questions (choice/score/noul) in one pass — the full engine |
| classify | assign the single best category (one choice) |
| score | rate on an ordinal scale, e.g. urgency (one score) |
| check | a yes/no/unknown gate for control flow (one noul) |
| triage | a ready-made decision set via a laya preset (triage / email / moderation / guard) |
| detect_language | script + language of a text (sub-ms, no model) |
| explain_routing | which checkpoint would answer, without running inference |
| list_models | the checkpoints available, default, device, offline status |
Three checkpoints, auto-routed per request: **english** (ModernBERT-large), **multilingual** (mmBERT, 100+ languages), **typed-decisions**.
Install
Python 3.10+. laya pulls in torch; the model checkpoints download **once** from Hugging Face (see below), after which it runs fully offline.
First run: fetch the model once
Downloads are **off by default** (nothing leaves your machine at runtime). Pre-fetch the checkpoints one time with network access:
Then run the server offline.
Configure
See docs/CLIENTS.md for Claude Code / Cursor / Codex / VS Code / Windsurf, and .env.example for every variable.
Example
Ask your agent to *"use laya to classify this ticket's department and whether it's a churn risk"*:
Safe by default
laya is read-only inference, so the guardrails (in src/mcp_laya/security.py) are about privacy and resource control, not write-gating:
- **Offline by default** — the model runs locally; nothing is sent anywhere. The one exception, the first-time checkpoint download, is disabled unless LAYA_ALLOW_DOWNLOAD=true.