System One Connector

Go CLI and single-binary MCP server giving Claude Code, Codex, Hermes and pi an evaluate tool backed by Jev, d1, CLM or Laya.

MCP serverbetaMIT

View on GitHub ↗

Overview

System One Connector (formerly typesafe-mcp) gives AI agents direct access to decision models through one evaluate tool. An agent asks a question such as 'is this urgent?' or 'which team owns this?' and gets back a number or an option it can use in an if statement, rather than a paragraph to interpret. It is a Go CLI and single-binary MCP server, and evaluate setup mcp finds Claude Code, Claude Desktop, Codex and Hermes and registers the tool with each; evaluate setup pi handles pi.

Jev via TypeSafe or OpenRouter is the default, but the same connector runs other System One models: self-hosted CLM or Laya by pointing TYPESAFE_BASE_URL at a local server, Liquid AI's hosted d1, and Respan's Span-01 for noul questions. Profiles let you switch models without rerunning setup. The README's pitch is that typed answers with probabilities let an agent proceed on confident answers and escalate unsure ones, and that Jev typically answers in under half a second. It is MIT licensed and installs on macOS and Linux with a script.

What the decision model decides

Type	Question	Answers
Yes / no	Does this convey urgency?	—
Choice	Which team owns this?	billing · technical · sales
Score	Rate on a scale	—

How it works

Install the binary and run evaluate setup mcp with a key
Agent calls evaluate with state and typed questions
Selected provider returns probabilities
Agent branches on thresholds or escalates

Strengths

One tool for several agents
Works with hosted and self-hosted decision models
Profiles for switching models

Caveats

Needs a provider key or local server
Span-01 supports noul only

Getting started

curl -fsSL https://raw.githubusercontent.com/itsmostafa/system-one-connector/main/install.sh | sh, then TYPESAFE_API_KEY=your-key evaluate setup mcp.

At a glance

Type: MCP server
Status: beta
Language: Go
Runtime: CLI + MCP server
Has code: yes
License: MIT

Use cases and integrations

Infra / SDKs / Integrations537

How curated lists describe it

Go CLI and single-binary MCP server, formerly typesafe-mcp, that gives Claude Code, Claude Desktop, Codex, Hermes, and pi typed decisions from Jev, d1, CLM, or Laya.
AnotiaWang/awesome-decision-models
System One MCP connector to evaluate anything fast and cheap.
wh000wh000/awesome-jev-live
Go CLI and single-binary MCP server exposing TypeSafe judgments to Claude Desktop, Claude Code, and Codex.
hellogumbo/awesome-jev
MCP connector that gives Claude Code, Claude Desktop, Codex, Hermes and pi an `evaluate` tool, pointed at Jev or at an open System One model you host.
MrJev/awesome-jev

Frequently asked questions

What does the evaluate tool do?

Lets an agent ask typed questions and get probabilities it can branch on.

Can I use open models?

Yes, self-hosted Laya or CLM via TYPESAFE_BASE_URL, or Liquid AI's d1.

Which agents are supported?

Claude Code, Claude Desktop, Codex, Hermes and pi.

Related projects

askjev
Unofficial hosted or local Jev MCP server giving agents a quick second opinion: it picks yes/no, scale or choice and returns probabilities.
MCP serverbetaTypeScript
swift-jev
Swift library and CLI for Jev with enum-typed Choice, Noul and Score questions, plus an agent skill for coding agents.
CLIproductionSwift
jev-mcp (blakestone-x)
Python MCP server wrapping Jev as classify, score, check, match and screen tools with confidence on every answer.
MCP serverbetaPython
openrouter-jev-mcp
Python decision gateway and stdio MCP server exposing Jev via OpenRouter's alpha Decisions endpoint to Claude Code, Codex and Cursor.
MCP serverbetaPython
prompt2jev
Agent skill and CLI that converts plain requirements, LLM prompts or prompt-calling code into a typed Jev request and runnable script.
Agent skillbetaPython
typesafe-cli (geilt)
Stdlib-only Python CLI plus agent skill for asking System One typed Choice, Score and Noul questions from the shell.
Agent skillbetaPython