AI
Your AI assistant tests with you.
Reflex works with Claude Code, Codex and OpenCode. Your assistant writes scripts, runs them on your agents and finds out why something failed. You stay in charge.
- Claude Code
- Codex
- OpenCode
Write and run
The assistant knows every Reflex library, writes a script, runs it on an agent and reads the result. A new test takes minutes.
-- Can a customer get a quote?
local browser = Play.start()
browser.navigate("https://www.verzekeraar.example/auto")
browser.locator("id=kenteken").fill("AB-123-C")
browser.locator("id=postcode").fill("3833 LD")
browser.locator('role=button[name="Get a quote"]').click()
Assert.contains(browser.locator("css=.premie").text(), "per month")
Find out why it failed
When something fails, the assistant looks at the same run you do: which step, which page, which message. It tries another selector and proposes a fix.
A person decides
A proposed fix is verified first and then approved by a person. You type passwords into Reflex yourself; the assistant only sees that they are set.
Within the user’s permissions
The assistant signs in as the user and can never do more than that user. Every action lands in the audit trail.
Frequently asked questions
How do I connect my AI tool?
One command in your repository: reflex setup ai. It installs the Reflex skills and the connection to your server for Claude Code, Codex or OpenCode.
Does our code or data go to an AI vendor?
Your AI tool reads what it needs, under your own agreement with that vendor. Secrets never enter the chat. The healer can optionally use an AI model you choose, with a daily budget.
Watch your AI assistant write and fix a test
We show it live in an introduction.