Share local recordings with an AI assistant

A filename is a reference, not an upload. Pasting /Users/you/Eavesy/Recordings/session/screen.mp4 into a browser chat does not give the service access to that file. Decide which evidence the assistant needs, then provide it through a supported input.

Start with readable text

Open a completed capture and copy the relevant brief or transcript. Include a focused question, such as “Which decision changed between these two passages?” Keep enough surrounding text for the assistant to understand who said what and when. Review names and numbers before relying on them.

If your recording has no final brief yet, say that processing is incomplete. Do not present a placeholder or preliminary caption as the finished transcript.

Add the visual evidence

For a UI problem, attach the screenshot that shows the state. If timing matters and the assistant accepts video, attach the recording. Supported file types and size limits belong to the receiving service and can change. If an upload fails, use a smaller relevant excerpt or screenshot rather than assuming the assistant can retrieve the local path.

Eavesy records the screen and can transcribe audio. A text transcript alone does not establish the visual meaning of each frame. A transcript saying “the button is disabled” is a speaker's observation until the image is inspected.

When a local agent can read files

A coding agent running on your Mac may read the capture folder with your permission. Supply the actual absolute path. Ask it to name the files it opened and distinguish text inspection from image inspection. Remote agents and cloud coding environments cannot necessarily access your Mac's filesystem.

When MCP is appropriate

Use the hosted Eavesy MCP server for account recordings and dictations. It can retrieve supported text and source evidence after authorization. It does not make all local screenshot, clipboard, or video files available. See local recording context.

Review before sharing

Remove unrelated private messages and credentials from any text or visual material you share. Sending an attachment to another assistant is a separate data transfer from keeping it locally in Eavesy. Once the assistant has the evidence, ask it to restate the observed problem and identify any missing input before acting.

A handoff you can paste

Task: explain the observed failure before suggesting a fix.
Source: the attached screenshot and the transcript excerpt below.
Observation: [what happened, including the starting state and trigger].
Expected result: [the behavior you expected].
Transcript excerpt: [the relevant words and source timestamp, if available].
Unknowns: [anything the recording did not establish].
First tell me which evidence you can actually inspect and what is missing.

Replace the bracketed fields with your observations. For a local coding agent, add the actual folder path. For a browser assistant, use attachments rather than expecting it to resolve that path.

Check the handoff before acting on the answer

1. Does the assistant identify the intended recording or screenshot?
2. Does its description match the visible state and the spoken passage?
3. Does it distinguish an observation from a proposed cause?
4. Does it name the source it could not open?
5. Does the next action address your original question?

If it gives a confident diagnosis without inspecting the required visual evidence, provide the missing input and ask it to reassess. A plausible explanation is not a substitute for access to the source.