Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

typesafe-jev-opencode

Local MCP server that adds a jev_decide tool to OpenCode agents, with a conservative confidence-gated policy for a developer agent.

ProjectbetaMIT
View on GitHub ↗

Overview

TypeSafe Jev for OpenCode is a local stdio MCP server that gives OpenCode agents a jev_decide tool supporting noul, choice and score questions. It uses Node's native fetch and the official MCP SDK, keeps the API key outside OpenCode configuration files, and is designed for a global install across projects.

The README pairs it with TypeSafe's official skill and includes a conservative Developer-agent policy for confidence-gated decisions. It stresses that Jev is not a conversational replacement for Gemini, Claude or GPT, but a way for an agent to route or gate work using typed answers and probabilities. It is MIT licensed, with documentation in English and Portuguese. The tool is exposed to the agent as typesafe-jev_jev_decide after restarting OpenCode, and documentation is available in English and Portuguese.

What the decision model decides

TypeQuestionAnswers
Yes / noAgent-posed yes/no question used to gate an action(inferred)yes · no
ChoiceAgent-posed routing choice(inferred)—

How it works

  1. Install the official TypeSafe skill for OpenCode.
  2. Clone and install the MCP server; store the key in a user env file.
  3. Register it in opencode.json.
  4. The agent calls jev_decide and follows the confidence-gated policy.

Strengths

  • Key kept outside OpenCode config.
  • Global setup works across projects.
  • Includes a conservative gating policy.

Caveats

  • Requires Node 20+, OpenCode and a TypeSafe key.
  • Manual configuration steps.

Getting started

git clone https://github.com/moisesfilho/typesafe-jev-opencode.git && npm install, then add the server to ~/.config/opencode/opencode.json

At a glance

Type
Project
Status
beta
Language
JavaScript
Runtime
MCP server (Node)
Has code
yes
License
MIT

Use cases and integrations

How curated lists describe it

  • Jev is not a conversational replacement for Gemini, Claude, or GPT. It evaluates application state against typed questions and returns structured answers and probabilities that an agent can use to route or gate work.

    logicrw/awesome-jev-projects

Frequently asked questions

What tool does it add?

jev_decide, exposed to the agent as typesafe-jev_jev_decide.

Where is the key stored?

In a user env file sourced by the launch command, not in OpenCode config.

Does it replace the agent's chat model?

No, it adds typed decisions for routing and gating.

Related projects

  • Jevbridge

    Zero-dependency Node library with MCP and ACP adapters adding Jev decisions and gated computer use beside any LLM agent.

    MCP serverbetaTypeScript
  • jev-mcp

    Unofficial, read-only MCP server exposing Jev as four advisory tools: decide, route, risk score and yes/no gate.

    MCP serverbetaTypeScript
  • jev-mcp

    MCP server giving any agent Jev's classify, score, yes/no check and allow/confirm/block gating of risky tool calls.

    MCP serverbetaJavaScript
  • jev-mod

    Claude Code mod that adds $.jev so other mods can ask typed Choice, Score and Noul questions, for example to deny destructive commands.

    ProjectbetaTypeScript
  • jevcore

    Jev decision layer for DeepSeek Harness, MCP hosts and plain Node, offline by default with disclosed egress and opt-in gates.

    MCP serverbetaTypeScript
  • jev-use

    Claude Code, Codex and pi plugin that hands non-writing agent steps to Jev and escalates writing or unsure steps to the LLM.

    PluginbetaJavaScript