Guardrails & safety — page 5

Probabilistic soft-constraint evaluation for volunteer scheduling using TypeSafe Jev, with deterministic application policy.

Prompt injection protection & tool-call guardrails for LangChain / LangGraph AI agents. Screens tool outputs, gates risky tool calls (allow / human review / block), powered by TypeSafe Jev.

Side-by-side showcase: traditional LLM vs TypeSafe Jev (System One) on stock order policy decisions

Minecraft (Paper 1.21) chat moderation plugin powered by JEV from TypeSafe

Multi-agent healthcare pipeline using Google ADK, stdio MCP tools, and TypeSafe Jev safety gates for instant emergency triage and clinical draft QC.

A provider-agnostic policy and authorization layer for AI agents and AI-native applications.

Safety-first job application automation with TypeSafe Jev and Browser Use

Hosted Jev policy workspaces with tenant isolation, finite evaluations and revocable aggregate reports

Stranger danger for AI agents: detect prompt injection in content your agent fetches, before it reads it. Runs on TypeSafe's Jev System One model by default; local and keyword detectors included. Benchmarked, failures published.

Classify page context, score configurable risks and apply advertiser-owned brand-safety policy.

Open-source developer harness and runtime safety toolkit for AI coding agents powered by TypeSafe AI / Jev

Self-improving agents that know what's noise: typed decisions instead of generated text, a policy that evolves from the agent's own mistakes, and a report of how much of the gain was selection luck.

Static code scanner (SAST linter) that finds insecure usage of Jev / TypeSafe guardrails in Python. Not a guardrail itself — build-time analysis, offline, no API key.

A safety guardrail that screens a user message against an assistant's policy before the assistant ever sees it.

Semantic MCP middleware that intercepts AI-agent tool calls and combines deterministic policy with TypeSafe Jev risk judgments before actions reach upstream MCP servers.

Minesweeper played by TypeSafe's Jev System One model: one P(mine) per hidden cell per move, code owns the policy.

Benchmark harness evaluating TypeSafe's Jev model on six public safety benchmarks against Shieldstral-1.0-3B and top SOTA guard models

Jev-powered skill router & security auditor for any AI agent (Codex, Claude Code, OpenCode, Hermes): ONE cheap decision per request tells the model WHICH skill to load; scans skill libraries for prompt injection & dangerous commands.

Snake auto-played by TypeSafe's Jev System One model: one typed decision call per tick, code owns legality and safety.

Sub-120ms smart LLM gateway & real-time optics dashboard using TypeSafe Jev System One. Defeats cache thrashing on coding agents (Goose, Cursor, VS Code) with Break-Even Cache Affinity and 0.60 Confidence-Gated Safety.

Go client for the TypeSafe System One API (Jev) — the first piece of sael, a content-safety classifier for an AI request relay

Support ticket triage: TypeSafe Jev judgments plus an explicit policy layer for priority and team routing, with a Next.js triage workbench.

Interactive smart home demo comparing local Laya (MLX) with Jev (TypeSafe API) using shared routing and safety rules.

An AI content moderation workflow powered by the [TypeSafe/Jev model](https://typesafe.ai).