Iris
VerifiedIris equips AI coding agents with direct visibility into running applications for reliable verification.
What is Iris?
Iris addresses the common issue where coding agents complete tasks but cannot confirm their own results. By exposing internal app events such as failed API responses or missing UI elements, it delivers precise feedback that guides fixes directly to the relevant source code. The system operates through four core actions: querying specific app states, performing interactions within an active session, streaming observations back to the agent, and issuing pass-fail verdicts supported by data. These steps integrate with existing workflows and reduce token usage compared to full page dumps. Developers retain oversight via a real-time panel while agents execute repeated checks after each edit. The approach complements browser automation tools by focusing on internal verification rather than driving the interface itself.
Key features
What you can use Iris for
Automated Verification of Agent Edits
Iris enables coding agents to check their own work inside running app sessions by monitoring changes and providing evidence-based assertions instead of relying on manual clicks or screenshots.
Detection of Silent Failures
The tool exposes hidden issues such as failed network calls, missing DOM elements, untriggered signals, and console errors that screenshots cannot reveal, mapping them back to specific component files.
Low-Token Interactive Testing
Using MCP-based Look, Act, Observe, and Assert operations, Iris supports targeted queries on real authenticated sessions with support for webhooks, websockets, and async jobs while keeping token usage minimal.
How to use Iris
- 1Install the dev SDK via npm
- 2Connect Iris to your localhost running app
- 3Invoke Look to ask targeted questions about elements
- 4Use Act to perform clicks or navigation in the session
- 5Observe network, DOM, console, and signal responses
- 6Assert pass or near-miss verdicts with evidence
Iris pricing
Pricing model: Open Source. Plan details are indicative — check the site for current prices.
Open Source
- MIT License
- npm i -D @syrin/iris
- Dev-only SDK
Editor's verdict
Pros
- +Deterministic verdicts with evidence instead of screenshots
- +Reduces verification time from minutes to seconds
- +Works with any LLM and requires no vision model
Cons
- –Dev-only SDK, not for production use
- –Complements rather than replaces full E2E tools like Playwright for CI
Our take: Iris is a solid coding & dev choice. It's valued for deterministic verdicts with evidence instead of screenshots and reduces verification time from minutes to seconds. The main trade-off is dev-only sdk, not for production use. A good pick if you want capable AI without a high upfront cost.
Frequently asked questions
It tracks DOM changes, network calls, route changes, console errors, animations, app signals, and regressions during real running sessions.
Summary
Iris is a solid coding & dev choice. It's valued for deterministic verdicts with evidence instead of screenshots and reduces verification time from minutes to seconds. The main trade-off is dev-only sdk, not for production use. A good pick if you want capable AI without a high upfront cost.
User reviews
Verified reviews from the community shape this tool's rating.
Loading reviews…
Iris alternatives
Similar coding & dev tools worth comparing.
