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

Jev-chooses-a-LLM

Jev Router for Cursor: Jev picks a COST, BALANCED or INTELLIGENCE tier per task, with soft routing in chat and hard routing via the SDK.

ProjectprototypeMIT
View on GitHub ↗

Overview

Jev-chooses-a-LLM is a router for Cursor in which Jev decides how much intelligence a task deserves, choosing among COST, BALANCED and INTELLIGENCE tiers, while Cursor executes and LLMs write and build. Its stated promise is to pick the cheapest available route likely to complete successfully under policy, not a mathematically optimal model. Jev is used only as a decision layer.

It works in two modes. In Cursor Chat with MCP (soft routing), the parent agent calls jev_route and may delegate to jev-cost, jev-balanced or jev-intelligence subagents, but the parent chat model is not hot-swapped. Through the Cursor SDK (hard routing), code runs route, validate, discover available models and create the agent with the resolved model, with dry-run and escalation support and a Cursor API key. Deterministic rules include confidence thresholds, and approval gates such as production deploy or financial transaction stop SDK runs unless explicitly approved. It is MIT licensed.

What the decision model decides

TypeQuestionAnswers
ChoiceHow much intelligence does this task deserve?COST · BALANCED · INTELLIGENCE

How it works

  1. Task arrives in Cursor chat or SDK
  2. Jev chooses a tier
  3. Chat: delegate to the tier's subagent; SDK: create agent with resolved model
  4. Approval gates stop sensitive runs unless approved

Strengths

  • Cheapest likely-successful route
  • Hard routing via SDK actually switches models
  • Approval gates for risky task types

Caveats

  • Chat mode cannot swap the parent model
  • SDK mode requires a Cursor API key

Getting started

Set TYPESAFE_API_KEY; for hard routing, `npx tsx sdk-runner/cli.ts --run "Fix the failing auth test"`.

At a glance

Type
Project
Status
prototype
Language
TypeScript
Runtime
Cursor MCP, Cursor SDK runner
Has code
yes
License
MIT

Use cases and integrations

How curated lists describe it

Frequently asked questions

What does Jev decide?

Which of three intelligence tiers a task needs.

Does chat mode switch models?

No, it delegates to tier subagents; only SDK mode hard-switches.

What are approval gates?

Task types like production deploys that stop SDK runs unless --approve is passed.

Related projects

  • ailerix

    Model router where Jev classifies each request's task family and code picks the cheapest capable model on a cost Pareto chain.

    ProjectbetaTypeScript
  • cursor-jev

    Cursor plugin whose hooks ask Jev whether subagent spawns, file reads, writes and shell commands are worth taking.

    PluginbetaJavaScript
  • Jevonian

    Local OpenAI/Anthropic-compatible proxy where one Jev call picks the model route and thinking level for each coding-agent turn.

    ProxybetaGo
  • auto-model-router

    Cost-, cache- and quota-aware LLM router that classifies each turn with a Jev-class model, prices every route, and checks cheap answers.

    ProjectprototypePython
  • Jevia

    Local-first, outcome-aware model router for coding harnesses that picks a capability tier per task with Jev and learns from history.

    ProjectprototypeRust
  • HiRoute

    Local-first routing engine for long-running agent tasks that uses a Jev decider to pick a model group per turn.

    PluginproductionRust, TypeScript