Qyrus testing, inside your coding agent
Your coding agent writes the code. Now it can test it.
Add QyrusAI Assure to Claude Code, Codex, Cursor or Antigravity. Your agent reads a work item, tests it across web, mobile and API, and returns evidence you can check.
Agent integrations
The problem
AI agents ship code faster than anyone can verify it.
When an agent writes the change, “it looks right” is not a test result. Teams need to know what was actually run, on what, and what happened.
Tests that vanish
One-off checks disappear when the chat ends, so the same work is repeated on the next change.
Gaps between surfaces
A feature works in the browser, but nobody checks the API behind it or the mobile app that uses it.
No evidence
It passed" with no screenshots, traces or failing step is hard to trust and hard to act on.
How it works
From a work item to test evidence
Give your agent a task. Assure connects it to Qyrus to create tests, run them, and report back.
01
Read the work item
One-off checks disappear when the chat ends, so the same work is repeated on the next change.
02
Generate scenarios
It drafts test scenarios, steps and data from the criteria.
03
Run real sessions
It runs the tests in a browser or on a device and captures what happens.
04
Save the test
The test is stored as a reusable asset, so the next change can rerun it.
05
Return evidence
You get results with screenshots, traces and the exact failing step.
A generated or queued test is not proof. The run ends with verified results, explicit failures, and the IDs of what was created.
Get started in three steps
This is the conversion block. It carries the three activation steps from the Phase 1 plan. Anchor ID: get-started.
Create your account
Sign up for a free QyrusAI account. This is where your tests, results and API keys live.
Already have an account? Sign in and continue to step 2
Get your API key
The plugin uses an API key to act on your account. You create it once and keep it private.
- Open your profile and choose User Token.
- Choose Generate Token and give it a name.
- Select the MCP scope and the service permissions your agent needs.
- Set an expiry date, then copy the key.
- Save it in a private file outside your code repository, for example env, as X-API-Key=<your key>.
Configure the plugin
Add the plugin to your agent, then point it at your key file. Setup takes one prompt.
Paste this prompt into your agent.
Set up QyrusAI Assure from https://github.com/QQyrus/qyrion
The agent checks and installs what it needs · It asks for the path to your credentials file · You confirm your Qyrus team · Reload your agent if it asks.
What your agent can do
Testing that
outlasts the chat
Four things your agent can do once Assure is connected.
Test across web, mobile and API
Turn tasks into saved tests
Rerun after every change
Reduce repeated reasoning
A saved test can replace an agent re-deciding the same steps on every run. We are measuring the difference in token use with an open experiment.
The Qyrus MCP server for reusable web and API tests, Qyrion for browser and device sessions, and the qyrusai Python SDK for test generation.
Try it with a prompt
Start with one of these
Paste any prompt into your agent after setup.
P1
Validate a story
Read story [ID] from my tracker. Write test scenarios from its acceptance criteria, run them against [app URL], and give me the evidence.
P2
Reproduce a bug
Reproduce bug [ID] on [app URL]. Show me the failing step, then save the check as a reusable test.
P3
Test a code change
Look at my latest changes. Find the tests they affect and run them on [build URL]. Summarise what passed and failed.
P4
Build reusable coverage
Test the checkout flow on [app URL] in the browser, then save it as a reusable test I can rerun after each release.
Connecting a work tracker needs its own sign-in with your agent. Replace the [brackets] with your own details.
Built with guardrails
You stay in control
Assure gives your agent the ability to test. It does not give it the authority to decide. These limits are built into how it works
Reading is allowed by default.
01Reading is allowed by default. Commenting, changing a ticket's status, publishing or deleting needs your explicit approval.
A passing run does not close a bug
02A passing run does not close a bug, merge a change or deploy a build. Those decisions stay with you.
Your API Key
03Your API key is scoped to the permissions you choose and has an expiry date you set.
Technical Debt
04Assure is not a background service. Work starts when you ask your agent to start it.
What is QyrusAI Assure?
A plugin that connects your AI coding agent to Qyrus, so the agent can create tests, run them on web and mobile, save them for reuse and report back with evidence.
Which AI agents does it work with?
Claude Code, Codex, Cursor and Antigravity.
Is it free?
Yes, there is a free tier.
What do I need to install?
The plugin. During setup it checks for Python and uv, and installs the Qyrion command-line tool and the qyrusai SDK in an isolated environment if they are missing.
Which systems does it run on?
macOS (Apple Silicon), Linux (x64) and Windows (x64).
Where do I keep my API key?
In a private file outside your code repository. Your agent asks for the file path during setup; you never paste the key into a prompt.
Can it read my tickets and stories?
Yes, through your agent’s connection to your work tracker. You sign in to the tracker separately. Reading is allowed by default; writing back, such as comments or status changes, needs your approval.
Does it run on its own in the background?
No. It runs when you ask your agent to start a task, or when a scheduler your agent supports starts it.
Give your agent the task. Get back the evidence.
Create a free account and
connect Assure in minutes.
The question isn’t whether to understand, change, and operate as one system. It’s how long you can afford not to.
