#typesafe

Let Claude code, chatgpt codex or control your own Chrome. Chrome extension + MCP server: Jev, TypeSafe's decision model, picks each click in ~0.5 s for a fraction of a cent. MIT, bring your own OpenRouter key.

Jev-first portable Agent Skill and optional MCP bridge for Claude Code, Codex, Cursor and compatible agents.

Research design for a Jev-native agent system:enable more options than jev provided with virtulization and paging, tool integration, external helper logits Top-k proposals with Jev-controlled fallback ,decision-aware hierarchical memory, and dependency-aware replanning.

Practical agent skills and examples for building with Jev. API setup, routing, ranking, and evidence checks.

AI life-and-civilization simulation: TypeSafe Jev makes every decision (typed, probabilistic, auditable); LLMs plan — OpenAI-compatible APIs, local models (Ollama, LM Studio), Claude Code, Codex.

TypeSafe Jev-powered intent gate for OpenCode: confirm intent before the agent dives into underspecified requests.

Claude reads the part of a large file that answers your question, not the whole file. A PreToolUse hook. Measured: -36% cost on the reads it fires on, no answer lost.

Inspectable Jev model and effort routing for Codex CLI and Claude Code, with local usage reports.

Risk-aware model routing for Claude Code, powered by TypeSafe's Jev

Agent memory that decides what to hydrate — local topology recall + TypeSafe Jev gates (not another dump-everything store).

Reward-hack radar for coding agents: structural denies + TypeSafe Jev System One sidecar for Claude Code & Cursor hooks

Pick the right LLM for each prompt with a calibrated decision model (TypeSafe Jev). CLI, Claude Code plugin and Pi extension.

jevmem saves the decisions, rules and failed approaches from your Claude Code chats to JEVMEM.md in your repo, and brings the relevant ones back next session.

AI-assisted email, calendar, tasks and documents: TypeSafe Jev gives typed, probability-backed judgments; LLMs draft (OpenAI-compatible APIs, local models via Ollama/LM Studio, Claude Code, Codex); you approve every action.

Claude Code skill: hand bulk judgment calls to Jev. 86% fewer Claude tokens on a 12-task benchmark, up to 20x faster. One-line npx install.

Style and quality linter for Norwegian and English text - MCP server for agents, with TypeSafe's Jev as the judgment layer. Reports findings, never authorship verdicts.