---
type: "post"
id: "zulali-s-1994-jeep-cherokee-kept-failing-california-759ee4"
slug: "zulali-s-1994-jeep-cherokee-kept-failing-california-759ee4"
title: "Zulali’s 1994 Jeep Cherokee kept failing California’s emissions test."
description: "A public repository describes xj-scanner, a laptop-based interface for reading a 1991–95 Jeep engine computer through a low-cost Arduino Nano, with Claude Code, Codex, or another coding agent helping the user operate it."
retrieval_nugget: "A public repository describes xj-scanner, a laptop-based interface for reading a 1991–95 Jeep engine computer through a low-cost Arduino Nano, with Claude Code, Codex, or another coding agent helping the user operate it."
published_at: "2026-11-17"
updated_at: "2026-11-17"
record_date: "2026-11-17"
date_kind: "published_at"
topics: ["claude","claude-code","codex"]
entities: ["xj-scanner","Claude Code","Codex","Arduino Nano","Jeep Cherokee XJ"]
related_patterns: []
novelty_label: "new_fact"
what_changed: "A Jeep owner used Claude-assisted analysis with a $3 Arduino Nano to diagnose a 1994 Cherokee’s emissions failure."
prior_refs: ["https://newruntime.com/posts/during-a-test-claude-sent-philadelphia-police-a-2aad16/"]
source_url: "https://github.com/MZULALI/xj-scanner"
source_urls: ["https://github.com/MZULALI/xj-scanner"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:zulali-s-1994-jeep-cherokee-kept-failing-california-759ee4"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/zulali-s-1994-jeep-cherokee-kept-failing-california-759ee4/","markdown":"https://newruntime.com/posts/zulali-s-1994-jeep-cherokee-kept-failing-california-759ee4.md","json":"https://newruntime.com/posts/zulali-s-1994-jeep-cherokee-kept-failing-california-759ee4.json"}
markdown_url: "https://newruntime.com/posts/zulali-s-1994-jeep-cherokee-kept-failing-california-759ee4.md"
---

# Zulali’s 1994 Jeep Cherokee kept failing California’s emissions test.

## Retrieval answer

A public repository describes xj-scanner, a laptop-based interface for reading a 1991–95 Jeep engine computer through a low-cost Arduino Nano, with Claude Code, Codex, or another coding agent helping the user operate it.

## Summary
A public repository describes xj-scanner, a laptop-based interface for reading a 1991–95 Jeep engine computer through a low-cost Arduino Nano, with Claude Code, Codex, or another coding agent helping the user operate it. The project matters as a concrete example of an agent-guided workflow that reaches beyond software, but its documented test is limited to one vehicle and it does not establish that the tool repaired an emissions problem.

## What happened
The xj-scanner repository says it can read engine-computer data from a 1991–95 Jeep using three wires connected to the diagnostic plug under the hood and a $3 Arduino Nano connected to a laptop over USB. The project describes reading fault codes and live engine data, logging a drive, and running the computer’s built-in tests. It was built and tested on a 1994 Jeep Cherokee XJ with a 4.0L engine and computer part number 56030117. The repository says other 1991–95 Chrysler and Jeep computers in the SBEC, OBD1 family should be close, while telling users to measure their own connector before connecting anything.

The README presents two ways to use the project. In the agent-guided path, a user clones the repository and opens it in Claude Code, Codex, or another coding agent, then asks for help hooking it up. The agent reads the repository’s `AGENTS.md` and guides the user step by step through software setup, breadboard wiring, identifying connector cavities with a multimeter, and taking an initial reading. The README says users can ask the agent to explain steps in simpler terms. The repository also includes step-by-step playbooks under `.claude/skills/` and documentation on the connector, wiring, protocol, parameters, and troubleshooting.

The manual route uses Python and Arduino tooling. The instructions create a Python virtual environment and install dependencies from `pc/requirements.txt`; a simulated Jeep mode (`DEMO`) is available to check the laptop side. Users install `arduino-cli`, install the AVR core, compile and upload the `arduino/xj_bridge` sketch, and run the scanner against an automatically detected serial port. The repository lists the scanner’s functions as fault-code reading, live data, logging, and actuator tests; helper scripts can summarize logs, sweep parameters, and check the computer’s part number.

The project’s central safety instruction is emphatic: connector cavity 6 carries 12 volts with the key on, and must never be connected to the Nano. The README warns that on its test car the cavity’s wire had the same black-and-white color as the ground wire, so color is not a safe guide. It directs users to check every cavity with a multimeter and consult `docs/connector.md` for expected measurements. The hardware list includes a 5V Nano, breadboard, jumper wires, three 1/4-inch male spade terminals, and a multimeter; the README estimates about $15 for the setup overall. The repository is GPLv3 and describes the project as a hobby effort used at the user’s own risk.

## What the archive adds
We have not covered this before; it opens a new thread.

## Trend signal
This does not yet fit a pattern New Runtime tracks. The repository offers an interesting instance of an agent helping a person interact with a physical system, but a single hobby project does not establish a durable product boundary, operating model, or broader shift.

## Why it matters
For builders, the useful idea is not that an agent can independently diagnose any old vehicle. It is that a repository can combine an understandable human-facing procedure, low-cost interface hardware, explicit checks, and agent-readable instructions so a user can carry out a specialized workflow with guidance. The division of labor is visible: the Arduino bridges a legacy diagnostic connector to USB; the Python program interprets and logs data; the coding agent can explain the setup and help the operator use it.

That structure may be relevant to teams connecting agents to equipment or other constrained environments. The project does not remove the need for domain knowledge or safe procedures: it makes one documented route more approachable while putting a critical electrical check in front of the connection. The source’s warning about cavity 6 also illustrates why guidance for agents should make dangerous steps and verification conditions explicit, rather than relying on wire colors or an agent’s general confidence.

## Limits and uncertainty
The repository says the tool was built and tested on one 1994 Jeep Cherokee XJ configuration. Its statement that other computers in the 1991–95 SBEC, OBD1 family should be close is not a report of testing those vehicles. The repository does not establish compatibility across models, connector variations, or hardware revisions. It also does not report independent validation of the scanner’s readings, results from actuator tests, or a measured emissions outcome. While the discovery headline frames the project around a California emissions failure, the repository’s description is of diagnostic access and tooling; it does not say the scanner found or fixed that vehicle’s problem.

## Skill to build now
When an agent is guiding work that crosses from software into hardware, make the procedure auditable before asking it to act. Separate the checklist into preparation, a simulated or otherwise safe first test, measurement of the actual connector, and connection only after the measurements match documented expectations. Put irreversible or hazardous conditions in an explicit stop instruction. In this project, cavity 6’s 12-volt warning and the multimeter check are the concrete examples to borrow; do not infer pin assignments from wire color.

## Reader takeaway
xj-scanner is a useful demonstration of an agent-guided, documented workflow for a specific legacy vehicle computer, not proof of a general automotive repair capability. Its most transferable lesson is to pair agent help with measured checks and explicit hazards—and to keep compatibility and repair claims bounded by what was actually tested.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Claude](https://newruntime.com/topics/claude.md): Explore the claude topic hub.
- topic hub: [Claude Code](https://newruntime.com/topics/claude-code.md): Explore the claude-code topic hub.
- topic hub: [Codex](https://newruntime.com/topics/codex.md): Explore the codex topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)
