Zight MCP Server
Stop describing the recording.
Let your assistant read it.
Paste a Zight link into Claude or any MCP-compatible tool and it reads what is actually behind that link — the screen, the transcript, the text visible on it, the console logs. The copy-and-paste step in the middle stops existing.
https://mcp.zight.com/mcp
Add it as a remote MCP server or custom connector
Trusted by 5m+ people who would rather show than tell
What is the Zight MCP server?
The Zight MCP server is a remote MCP server that lets AI assistants read the context behind a Zight link: the screen recording, its transcript, the text visible on screen, and the console and network logs that came with a request video. Connect it at mcp.zight.com/mcp.
What it can read
Four kinds of context, from one link
An assistant that can only read your summary of a recording is reasoning about your summary. This is the recording itself, and everything Zight captured alongside it.
Screen recordings
The walkthrough, the bug report, or the customer’s own recording — as the source, not as a summary of it.
Transcripts and OCR
What was said, and the text that was visible on screen. No copying error messages out of a video by hand.
Console and network logs
The technical evidence that came with the recording, so debugging starts from what actually happened.
Request videos
The recordings people sent you through a request link, with the same context attached.
The transcript and on-screen text an assistant reads are what Zight AI produces from the recording; the logs and customer recordings come from request links.
What people do with it
Three jobs that stop needing a copy-paste step
Debug from the evidence, not the retelling
Give your assistant the link to a customer’s request video and it works from the recording, the transcript and the console logs together — instead of from your paraphrase of what they said went wrong.
Turn a walkthrough into the follow-up
Point it at a recording and ask for the summary, the action items, or the handoff note. The assistant reads what was said and what was on screen, so the output is about the actual work.
Stop pasting context into chat
The copy-the-transcript-into-a-chatbot step disappears. The assistant fetches what it needs from the link, which means it is working from the whole recording rather than the part you remembered to paste.
FAQ
Questions people ask first
A remote MCP server that lets an AI assistant read the context behind a Zight link — the recording, its transcript, the text visible on screen, and the console and network logs when a request video carried them.
Any app that supports remote MCP servers or custom connectors, which includes Claude. What a given app can do with the context depends on its own MCP implementation and the permissions you grant it.
Use https://mcp.zight.com/mcp as the remote MCP server URL. Add it as a remote MCP server or a custom connector, depending on what your tool calls it — the Help Center has the step-by-step for each.
No. The assistant reads the context behind the Zight links you give it, and it inherits the permissions of the account that connected the server. Zight is SOC 2 Type II certified and your content is not used to train models.
It helps. The transcript and the on-screen text an assistant reads are what Zight AI produces from the recording, so the two work together — the MCP server is how that context leaves Zight.
Give your assistant something real to work from.
Record it in Zight, paste the link into your AI tool, and let it read the whole thing.