- Category
- Developer Tools
- Rank
- No. 2231Tools index
Previous survey · No. 2122 ·
- Type
- TOOL
- Use case
- Models: Train & Run
- Interfaces
- CLI
- GitHub
- 2 stars
- Latest release
- 0.4.1
- Date
About
decide is a command-line tool that sends a question, along with a fixed set of possible answers and optional context, to a Jev decision model and returns a single calibrated answer with no prose to parse. It supports single or multiple-choice questions, graded severity levels with per-level descriptions, and multiple named questions in one call, aimed at scripts and coding agents that need fast, cheap judgment calls instead of full LLM reasoning.
At a glance
decide asks a model a bounded question and returns a yes/no answer, a choice or a rating for a shell script to consume. It is worth trying for low-stakes routing where you can define the allowed outcomes and send ambiguous cases to a person.
Good fit
Developers adding a model-based classification step to scripts or agent workflows.
Consider something else when
Skip it when a deterministic rule works or a wrong answer could trigger a consequential action without review.
Start here
Route three synthetic support tickets into two queues and set a human-review fallback. Include one ambiguous ticket and one simulated provider failure. Verify that the script handles fallback and exit codes before connecting real input.
Cost and requirements
The CLI is Apache-2.0. Its documented setup needs a configured decision model and API key. Model usage may be billed by the provider.
Limits to know
The configured provider receives the question and any supplied context. Model confidence is a routing signal. The CLI distinguishes undecided, setup and remote errors, and a configured fallback can turn a remote failure into a normal output.
Sources checked 2026-11-05. Source-based assessment.
Why it made the leaderboard
Gives scripts and agent skills a fast, cheap way to make calibrated judgment calls (routing, urgency, classification) without invoking a full LLM for every decision.
Intel on decide
- How to Decide What Your AI Agent Should Remember - Wojciech Pluta, Oracle
- Default to serverless for AI agents, then let production data decide what moves
- The Window Is in the Reasoning: When LLM Agents Decide to Follow Prompt Injections
- Mercury Decide adds image input on OpenRouter
- Mercury Decide adds a zero-data-retention endpoint on OpenRouter
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.