What is Jev Codex Factory?

A multi-agent coding factory for OpenAI Codex that turns one engineering request into a routed, parallel, verified, resumable swarm of coding agents: Jev decides how to execute the task, Codex agents do the work, and verification gates every change.

Does Jev write the code?

No. The README is explicit: Jev does not write the code, it decides how the code should be written. A first Jev shape call asks whether the task is single-agent or a DAG, and a router then picks the model and effort lane.

What happens when verification fails?

A second Jev call classifies the failure into retry, escalate, block or needs human, rather than the workflow blindly retrying. Blocked work can then be resumed later instead of being lost or restarted.

What kinds of jobs is it for?

Jobs that get messy as a single model call: independent subtasks, different reasoning difficulty, missing dependencies, flaky verification, parallel file edits, partial failures and interrupted sessions. Small tasks do not need the factory.

How is it different from a typical coding agent?

The README contrasts one-model-for-everything and a hope-tests-pass loop with per-task model and effort routing, parallel agent execution, verification on every change, and classified failure handling instead of blind retries.

What license and status does it have?

Apache-2.0, with a CI workflow and a v0.1.0 release badge in the README. The GitHub metadata shows it was created on 2026-09-27 and has one star, so treat it as an early-stage project rather than a proven tool.

What is the shape call in this system?

The first Jev decision, labelled shape #0, asks one thing: is this task single or should it be decomposed into a DAG. A single answer routes to one Codex agent with a chosen model and effort; decompose produces the task DAG for parallel workers.

Jev directory

CommunityJev Agents & MCP

Jev Codex Factory

A Jev-powered multi-agent coding factory for OpenAI Codex: one engineering request is routed to a model and effort lane or decomposed into a task DAG, workers run in parallel, verification gates every change, and failures get classified instead of blindly retried.

See highlights Open github.com

Stars: 1
Language: Python
License: APACHE-2.0
Reading time: 2 min
Published by: Community
Author: vyetgokyra
Last verified: Oct 2026

vyetgokyra/jev-codex-factory

User repository on GitHub

View on GitHub

Route smarter. Code in parallel. Resume what breaks. A Jev-powered multi-agent factory for Codex.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity catalog-import logicrw codex multi-agent

Works forJev for Routing Jev for Code Jev for Decision gating

Used inJev for Software & DevTools

Languages: Python
Capabilities: Open source

Highlights

Jev decides how the code should be written rather than writing it: a shape call first asks single agent or decompose, then a router picks model and effort.
Decomposition produces a task DAG with parallel workers A, B and C, each working through Codex on its own part.
Verification gates every change; a failure goes to a second Jev call that classifies it as retry, escalate, block or needs human.
Blocked work can be resumed later, and the README targets jobs with independent subtasks, flaky verification, partial failures and interrupted sessions.
Apache-2.0 licensed, with CI and a v0.1.0 release badge in the README.

Watch out

Copy reviewed against the repo README on 2026-10-15; the README is truncated in the digest after its comparison table, so installation, configuration and any benchmark numbers are not covered here. The repo metadata shows it was created on 2026-09-27 with one star, so it is new and unproven.

Questions about this entry

Short answers drawn from the linked project or page.

From the Jev guides

Original write-ups on what this entry is for.

MCP tool routing: Jev in front of the tool call

Oct 16, 2026

Tool menus can outgrow the context that reads them. The catalog's Jev-based routers and gates narrow the list: benchmarked tool search, per-tool scoring, fail-closed approvals, and WebMCP in the browser.

Jev Agents & MCP#mcp#tool-routing#agents

Read article

Approval policies for coding agents: hooks, guards, and gateways