ajx.

Agent experience, observed.

Your next user
is an agent.

Find out where it gets stuck.

AJX gives a fresh agent a real task in your product, records the journey, and turns friction into improvements your team can act on.

Get started with your agent ↗See the evidence ↓

Open source. Local-first. Your harness, your models.

field-notes / csv-exportFictional sample
The task

“Export 100 records to a CSV file.”

⑂
Two commands. No clear choice.

export or dump?

E-002
!
The quick start doesn’t run.

error: unknown flag --out

E-003
◷
Working… or stuck?

47 seconds without progress.

E-006
✓
The task succeeds.

100 rows. CSV verified.

E-008

Passed the task.
Still found 3 things to improve.

↗
Bring the tools you already use
Codex·Claude Code·Kiro CLI·Agent Skills-compatible hosts

01 / Start small

One prompt.
One real task.
A better first impression.

Open your project in a skill-capable agent and paste this prompt. It handles the AJX installation and guides your first run.

  1. 1
    Bring your project and an agent

    Use macOS or Linux, with a supported worker CLI installed and signed in.

  2. 2
    Agree on a small task

    Review the success check, access, and expected model costs before anything runs.

  3. 3
    Open the results

    Get prioritized asks, a visual journey, and the evidence behind both.

Paste this into your agent 
Install the latest stable AJX CLI and skill by following
https://raw.githubusercontent.com/royosherove/ajx/main/docs/AGENT-QUICKSTART.md
Then load the installed AJX skill.

I want to use AJX to test jq.
Starting point: https://jqlang.org/manual/
Task: Filter a sample JSON inventory to in-stock items and export their names and prices as CSV.

Prepare any tools, generated sample data, and local test server needed in a temporary workspace.
Use the AJX skill to run one trial with a fresh agent, verify the result, and show me the report and journey.
Try an example. No project needed.CLI + skill installed together
Prefer a terminal? Install the CLI + skill yourself. 
bash -o pipefail -c 'curl -fsSL https://github.com/royosherove/ajx/releases/latest/download/install.sh | bash' && export PATH="$HOME/.local/bin:$PATH"

macOS or Linux. The installer includes the skill. Your agent can read it directly; start a fresh session if the host needs one to discover /ajx. No GitHub login required.

Existing skill? Get the complete replacement command.

The native installer includes the Go CLI and the AJX skill.
No Go toolchain or GitHub login needed.

02 / Look past “it worked”

The useful part is
what happened along the way.

A pass tells you the task finished. The journey tells you how much help your product needed to get there.

Demo Export CLI

Make the quick-start
export command executable.

First thing to fix
What the agent encountered

The documented --out flag failed. Command help showed --output, which worked.

A concrete product change

Use the supported flag in the quick start. Include a copyable CSV example that runs in a fresh workspace.

Evidence, not guessworkE-003 → E-004 → E-0051 failed call · 1 recoveryInspect the full ask ↗
Fictional example · not a measured product resultASK-001 / 3 asks
⑂
Walk the journey

Forks, blocks, waits, and breakthroughs.

↗▦
Compare the matrix

Versions, harnesses, outcomes, and asks.

↗

03 / A repeatable way to listen

Your agent sets it up.
A fresh agent tries it out.

01 — Brief

Define the experience.

Choose the task, harness, model, environment, and a success check you can verify.

02 — Observe

Let the attempt unfold.

A fresh worker tries the task. AJX records its actions and preserves the output artifacts.

03 — Understand

Turn friction into asks.

Review verification, trace-backed findings, and the journey. Choose what to improve next.

When you're ready

Repeat the task across harnesses and models, or use ajx ci with acceptance checks in your pipeline.

Read the docs ↗