Dimension scores are derived from public data and fields; weighted into the composite. Reference only.
Iris is an open-source screen recording and AI context extraction tool. Rather than positioning itself as a traditional Loom-style screen-sharing product, it is designed to help AI Agents “understand” human screen activity. It can capture screen video, microphone audio, and browser behavior via a browser extension, CLI, or uploaded video, then convert recordings into timelines, actions, transcripts, URLs, code snippets, tags, and frame-level metadata for both humans and Agents to query.
Its AI capabilities primarily rely on Claude for visual frame analysis, extracting OCR text, UI states, active applications, and screen events. Deepgram is also used for transcription and processing. Integrations are a strong point: Iris supports an MCP Server for connecting to MCP-compatible Agents such as Claude Desktop and Cursor, and also provides a REST API, TypeScript SDK, and CLI. Typical use cases include QA bug reproduction, documenting development workflows, Agent-assisted testing, and building team knowledge bases.
The free plan requires no credit card and is available permanently, but the limits are significant: 5 active recordings, a maximum of 2 minutes per recording, 10 uploads per 24 hours, and 100 analyzed frames per recording. Pro costs $20/month and includes 30 hours/month, recordings up to 4 hours long, and 600 analyzed frames. Team costs $80/month/workspace and includes 200 hours/month, recordings up to 8 hours long, 1200 analyzed frames, unlimited members, and Google/GitHub SSO. Enterprise pricing requires contacting sales.
The advantages are that Iris is open source, can be self-hosted with Docker, offers comprehensive developer integrations, and produces highly structured output, making it well suited as contextual input for Agents. On the data side, the company states that it does not use customer content to train models, and users retain ownership of their content. The downsides are that the free tier feels more like a trial; AI processing depends on Anthropic and Deepgram; the terms still contain a TODO around jurisdiction, suggesting limited compliance maturity information; and there is no visible Chinese interface, Chinese documentation, or Chinese-language customer support.
Iris is best suited for AI engineers, Agent developers, QA teams, and technical teams that need to turn operational workflows into machine-readable knowledge. Access from China is not discussed in the main documentation, and because its key capabilities depend on Claude/Deepgram, real-world usage may be affected by network access, API keys, and USD payments. For domestic deployment in China, it may be worth looking at the self-hosted option or evaluating alternatives such as Loom, Tella, OpenReplay, and Sentry Session Replay.
⚠ This review is compiled from public sources and does not constitute a purchase recommendation. Verify all facts on the vendor's official site. Verify on getiris.xyz official site.
getiris.xyz is an Unknown AI Apps provider. TG4G tracks its product information, an overall rating of 8.0/10, and a China-accessibility score of China direct-connect friendly. Click "Visit Official Site" to reach getiris.xyz directly.