Agent
A judgment at the decision point
Use this when an agent already has a main model and needs one structured judgment at a decision point. Jev does not replace the agent's main model, does not run the agent's tools, and does not write a long rationale. Preset trials that ask you to log in, and the keys for those calls, stay on the Jev community site: the trial on jevai.org and the key page on jevai.org.
- The main model stays in charge
- Three ways in, none of them hosted here
- Keys never land on this hub
- Read the shape, then build a request locally
The main model stays in charge
The agent's own model still plans, writes, and decides which tool to pick up. Jev is asked for one structured judgment, and the agent continues from that result.
Jev does not replace the agent's main model. It does not run the agent's tools. It does not write a long rationale.
Three ways in, none of them hosted here
The community site offers three connections. The preset trial at jevai.org/agent asks you to log in. MCP exposes the same judgments as remote tools. A Skill is public Markdown that names which tool to call and which fields to send.
JEV Hub explains that split. It does not run their presets. Read the six remote jev tools or the Skill files the agent reads first.
Keys never land on this hub
A personal key is issued only at jevai.org/agent/keys. JEV Hub does not issue keys, does not store them, does not run the community presets, and does not forward calls to their API.
Do not write a key into a Skill file, a project file, or a web page. This page has no key form and does not save a key.
Read the shape, then build a request locally
The primitives under those judgments are Choice, Score, and Noul. The request builder validates JSON and does not call an API.
This hub has no trial runner and no progress animation. To run a preset while logged in, open the community trial.