What exactly leaves my machine when I enable the live provider?

With the mock provider and both gates off, nothing leaves: the startup log reports egress off. Enabling live mode names each transmission, with tools sending only called arguments and gates sending tool payloads after redaction.

Why default to an offline mock instead of just calling Jev directly?

Because earlier Jev wiring shipped prompts and file contents to third parties while enabled by default, so this project keeps the live path behind explicit provider and credential settings plus opt-in gates.

Should I route through TypeSafe directly or through OpenRouter instead?

Both return the same typed answers through the same SDK, but TypeSafe direct sends state to TypeSafe while OpenRouter sends it to OpenRouter with different retention, and only the OpenRouter route reports cost.

Can the model under judgment widen or disable the gates watching it?

It cannot. No tool exposes gate configuration, disabled gates register no event listener at all, and an undecided judgment resolves through explicit config defaulting to ask.

What are the three packages, and which do I pick?

jevcore is the decision layer with no harness dependency, for a plain script or service. jevcore-dsh is the DeepSeek Harness plugin with one service, three tools, and two opt-in gates. jevcore-mcp exposes the same three tools over stdio for any other MCP host.

What Node versions does each package need?

jevcore-dsh tracks the harness and needs Node ^22.19.0 or >=24.0.0, while jevcore and jevcore-mcp need Node 20 or newer.

What does redaction actually remove before sending?

Two passes: values under sensitive field names such as password, token, apiKey, and authorization are replaced wholesale, and secret-shaped strings in free text are pattern-matched, including Bearer tokens, sk- keys, ts_live_ keys, AWS/GitHub/Google key shapes, JWTs, and private-key headers. It is best-effort, not a guarantee.

CommunityJev Agents & MCP

Jevcore

Decision layer exposing Jev's typed yes-no, choice, and score answers to DeepSeek Harness and MCP hosts, offline by default with every network transmission disclosed per feature.

See highlights

Open github.com

Stars: 106
Language: TypeScript
License: APACHE-2.0
Published by: Community
Author: PerryLink
Last verified: Oct 2026

PerryLink/jevcore

User repository on GitHub

View on GitHub

TypeSafe Jev for DeepSeek Harness, the Model Context Protocol, and plain Node: typed judgments instead of prose, offline by default.

Forks: 1
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

About this project

TypeSafe Jev for DeepSeek Harness and any other MCP host. Jev is not a chat model. It answers typed questions — noul (yes/no), choice, score — and returns calibrated probabilities. It does not write prose, and asking it to is a category error. This project gives an agent exactly that surface, and nothing more. Offline by default. Egress disclosed. Nothing default-on. 📖 Ecosystem knowledge base — measured data, not marketing: plugin development guide · plugin-selection data · maintenance criteria.

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

Three surfaces sharing one core: plain-script decisions, a harness plugin, and a stdio MCP server.
Offline mock by default with per-feature egress report printed at startup for audit.
Two opt-in gates for safety and context that stay fully unregistered while disabled.
Live routes via TypeSafe or OpenRouter with key references, never inline secrets.

Watch out

Redaction is best-effort pattern and field-name matching that cannot catch an unrecognized secret in free text, so workloads that cannot tolerate any leak should stay on the offline mock. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

JevLoop

Runnable agent harness that moves pick, score and yes-no loop choices to Jev or Laya in milliseconds while reserving the language model for writing, compiling DECISION.md into executable questions.