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

decide-mcp

Local MCP server that lets an agent delegate a decision to Jev and get a recommended choice with percentages.

MCP serverbeta
View on GitHub ↗

Overview

decide-mcp is a local stdio MCP server with a decide tool: the agent sends a decision question, a context object and a list of choices, and receives a recommended choice and percentages for each option. The agent then decides what to do next. By default it calls TypeSafe's Jev through Vercel's AI Gateway using the AI SDK's experimental evaluate function, so it needs only an AI Gateway key.

It supports custom decision policies, multiple bias profiles, automatic routing between profiles and per-profile tools, configured by a JSON file or inline JSON. Credentials stay in environment variables referenced by name. It runs locally and is not a hosted MCP endpoint. Useful for agents that need a quick, structured second opinion on questions like whether to ship a change now.

What the decision model decides

TypeQuestionAnswers
ChoiceShould we release this change now?(inferred)ship · wait

How it works

  1. Agent calls decide with decision, context and choices
  2. Server applies the selected policy or bias profile
  3. Jev returns probabilities over the choices
  4. Agent receives a recommendation and percentages and acts

Strengths

  • One tool for arbitrary bounded decisions
  • Configurable bias profiles and routing
  • Runs via npx with no config file by default

Caveats

  • Requires an AI Gateway key
  • stdio only; other transports need extra integration
  • Early version (0.1.0)

Getting started

Add an MCP server entry running npx -y decide-mcp@0.1.0 with AI_GATEWAY_API_KEY in its env.

At a glance

Type
MCP server
Status
beta
Language
TypeScript
Runtime
MCP server (Node 22+)
Has code
yes

Use cases and integrations

How curated lists describe it

  • MCP ecosystem: configurable decision server with percentage scores and bias-profile routing on top of Jev.

    yibie/awesome-jev

Frequently asked questions

What does decide-mcp return?

A recommended choice plus percentages for each option.

Which model does it use?

typesafe-ai/jev through AI Gateway by default; other providers can be configured.

Is it hosted?

No, it runs locally as a stdio MCP server.

Related projects

  • a0-typesafe-ai

    Agent Zero plugin adding a typesafe_query tool that asks Jev Choice, Noul and Score questions and shows probability cards in chat.

    ProjectbetaPython
  • JevX

    CLI and MCP tool that finds hard-coded judgment calls in your code, scores how well Jev fits, and replaces strong fits with a fallback.

    ProjectbetaTypeScript
  • decisions-judge-mcp★ 5

    MCP server exposing a judge tool for typed noul, choice and score questions over JSON state, backed by Jev.

    MCP serverbetaTypeScript
  • jev-me

    Agent skill that grills a plan as a design-tree interview, using Jev for typed judgments along the way.

    Demoprototype
  • Building with Jev★ 155

    Agent skill for writing and improving programs that call Jev: question design, state structure, thresholds and diagnosing wrong answers.

    Agent skillproduction
  • jev-mcp (freepik-company)

    Go MCP server giving any client typed Jev decisions via OpenRouter or TypeSafe, with validated answers and no paid retries.

    MCP serverproductionGo