Upload and analyze a local file via EnriProxy (server-side extraction + model analysis).

When to use:

• Large PDFs (many pages) or scanned PDFs where client-side Read may truncate or miss content.

• Video/audio or other binary media your client cannot Read.

• Audio files in common formats (mp3, wav, flac, m4a, aac, ogg/oga, opus, wma, weba, mka, aiff/aif/aifc, caf, m4b/m4r, mp1/mp2/mpa/mpga).

• HEIC/AVIF/TIFF/APNG/SVG/Office docs when your client Read is unreliable.

• Very large files where resumable uploads are required (up to 4GB).

• Large PDFs/videos: set analysis_mode to 'multipass' for better coverage (auto prefers multipass for PDFs > 20 pages).

• For time-specific video questions (e.g., "what happens at 12:34?"), set video.clip_start_seconds and video.clip_duration_seconds.

Rules:

• Use path for one file, or paths for multiple images (UI screenshots/photo sets).

• path/paths are absolute paths on the machine running this MCP server (the client).

• Requires a valid EnriProxy API key (env ENRIPROXY_API_KEY, sent as Authorization: Bearer ...).

• Prefer the client's native Read tool only for small/simple text/PDF/common images when it works; prefer this tool for large PDFs.

• Answer strictly from the tool output; if frames/transcript are missing, say so.

• Video: frames + transcript belong to the SAME video timeline (not unrelated images).

• Animated GIF/WebP/APNG/SVG inputs are converted into representative key frames.

• Set language (e.g., 'es') to match the user request and avoid language drift.