Capture, then mark the point
Capture an area, stitch a scrolling image and annotate; optionally freeze the screen first.
A workspace for your clipboard and collected material
From captures and clipboard collection to text recognition, a built-in AI assistant and LAN transfers, keep scattered material easy to find and ready to use on another device.
ACTUAL APP · CHINESE UI

01 / CAPTURE & COLLECT
Capture, annotate and keep copied text. Enable automatic collection, or record several items and copy them together. File references save locations; back up files separately to keep copies.
FAQ: File references and backup copiesACTUAL APP · CHINESE UI
Circle key areas and add notes while keeping the original and editable annotations locally.

Capture an area, stitch a scrolling image and annotate; optionally freeze the screen first.
Keep text, images and file locations by type. Enable automatic collection when you need it.
Start recording to gather from chosen sources, then copy the text or local paths together.
ANNOTATE & ASK AI
Capture the issue, mark what matters with arrows and notes, then paste the annotated image path into your AI terminal. Keep the original and annotations locally to revisit and edit later.
/demo/feedback-annotated.pngExample path/demo/feedback-annotated.pngUse this annotated image to fix the button alignment at A.
Requires a terminal that can read local images
Five accent colors with Liquid glass or Classic styling; browse the three record types together or separately, reorder entries, or switch workspaces automatically.
FAQ: Appearance and workspace settings02 / TAG & ORGANIZE
Put captures, text and file references for one task into a single group, viewed together or by type. Pin what you need next and unpin it when done; the records and group remain.
ACTUAL APP · CHINESE UI
Pin the item to handle first in Today’s to-dos, then unpin it when finished while keeping the group intact.

Enable source identification to tag new captures and auto-collected content by app. App names are the default; your assigned tags take priority.
New captures go to Image processing with tags for the identified main app; unclear sources stay untagged.
With text auto-collection enabled, new copies enter Pasted content with their identified app tags.
Add tags yourself, filter records, and use tags to choose what to sync or share with AI.
Handy is a separate, free, open-source speech-to-text app; enable transcript collection to auto-tag new transcripts (recordings are never read).
FAQ: Collecting Handy transcriptsCopy three work notes and group them as Today’s to-dos. Pin the item you need next, then unpin it yourself after handling it. The text and group stay available.
Source detection and text collection on
All 3 records stay in the group
Edit pasted text directly; enable version saving to switch back anytime. Copy, search, sync and AI reads use the current content.
FAQ: Text editing and saved versionsAdd a short reminder without changing the original. Notes stay local by default; sync and AI reading need separate permission.
FAQ: Saving and sharing notes03 / TEXT RECOGNITION
Extract text with offline OCR, or connect a model for tables and formulas; proofread, copy or export the result.
ACTUAL APP · CHINESE UI
View the original image beside offline OCR text so you can proofread, copy or export right away.

ACTUAL APP · CHINESE UI
Connect a vision model to turn document images into text, tables and formulas ready to preview or export.

Built-in OCR works offline, with no model service to configure. Copy or correct the text, then search it to find the original image.
Connect a model that accepts images to recognize tables and formulas, then preview or export the result. Images go to your configured service; proofread the result against the original.
Batch-recognize and proofread PDFs, then export plain text or a searchable PDF.
Keep screenshots, copied text and file locations in their workspaces. When you need words from an image, run OCR and copy the result into your work.
In this example: collect with shortcuts · run OCR manually
Receiving OCR starts on: when OCR is enabled with an available engine, new incoming images join the queue automatically. Online engines send images to the selected service.
FAQ: Received images and automatic OCR04 / AI ASSISTANT
Chat inside SanFrame to find and organize permitted material, or ask how to use any feature. Off by default. Provide your own DeepSeek API key and keep SanFrame running; questions and permitted material read on demand go to the online service and may incur API charges.
ACTUAL APP · CHINESE UI
Keep separate conversations to find or organize permitted records, or switch to Guide only for app help.

Find text and organize groups and tags within your permissions; archive and restore conversations to continue anytime.
Switch to Guide only to search guides and permitted settings state and jump to settings, without reading or editing records.
Choose a vision-capable model and enable image reading separately; checking the permission does not itself upload images.
This example uses the built-in assistant to group two reading notes. Add your own DeepSeek API key, enable the assistant, then choose the records and actions it may use.
Group the reading notes and add the Shared reading tag.
Use allowed actionsBring essays and blank notes.
Keep a passage and your thoughts.
This example allows two ordinary text records and the required actions only. Privacy restrictions and other exclusions still apply.
Undo when recoverable. Sent content cannot be recalled.
These are illustrated review and undo entries, not live controls. Later edits or removed data may prevent undo.
05 / AI ACCESS & PRIVACY
Set record scope and action permissions in AI access on the top toolbar, or connect an external MCP-capable AI app. The built-in assistant and external AI share permissions with independent switches (both off by default). Keep SanFrame running during use. When an external app uses a remote model, permitted content may go to that service; fees depend on the provider.
ACTUAL APP · CHINESE UI
Filter by date, attributes, source and tags; block private records entirely or share redacted text only.

Enabled workspaces allow excerpts; full text, images, notes and actions need separate permission. Allowing images does not itself upload them, and mixed-type group relationships remain manually managed.
FAQ: Record scope and reading permissionsMark private records yourself, or check with rules or a downloaded local model (images need OCR first; checks can miss or misclassify content).
FAQ: Local privacy checksPrivate records are blocked by default. You can opt to share redacted text only.
Mark private content locally, select dates and material, let AI organize ordinary notes, and review what changed.
Contact details · Tagged
Local check → Private tag
This example enables local checks with rules, AstrLink Guard, or both. Download and enable the model, then choose detection types as needed. Synthetic samples do not demonstrate an actual model result or accuracy.
Copy MCP config → Connect AI
The user allows today's pasted text and separately enables creating groups, adding tags and favorites. Other workspaces and dates are excluded. Private records cannot be organized, even when redacted text is allowed.
Group today's design notes, tag them Design and favorite the color reference.
Changes recorded · Review or undo
Only the two ordinary notes are organized; the private record is unchanged. Authorized actions execute directly. Later edits may block undo, and text given to an external AI cannot be recalled.
Operation history distinguishes the built-in assistant and external AI so you can review and undo recoverable changes; sent content cannot be recalled.
FAQ: Operation history and undoMy workflow: let terminal AI organize and pin via MCP06 / DEVICE TRANSFER
Pair computers once; SanFrame reconnects on the same local network while running. Send text, images and files to Android phones running SanFrame Transfer too.
ACTUAL APP · CHINESE UI
Type and send text once connected; images and files have their own sending controls too.

Send text, photos, files or a whole group to your chosen devices. Enable text copying on a receiving computer to paste incoming text into your work.
Copy received text · Enabled before receiving(Off by default)Original text to clipboard · Line breaks kept · You paste
Each incoming update replaces the text. It stays in Device transfer by default, without saving to history. This example enables Copy received text on a receiving computer beforehand, then manually pastes the original into a note. This does not make other content types copy automatically.
07 / SYNC & KEEP
While connected on the same local network, sync new content automatically or send after favoriting, tagging or grouping. Older records need separate permission and an explicit choice; reconnecting does not backfill history or offline content.
ACTUAL APP · CHINESE UI
Pick what the group shares, then choose whether this computer sends new items or waits for a mark.

Choose shared content and how this computer sends it. Keep new material on another computer, or send after adding selected tags, favoriting or grouping.
Same LAN · Connected · New-content sending here
Updates saved
Enabled here
Free to download and use. No account needed.
Windows and macOS provide the full desktop tools. The Android app shares text, images and files with your computer.
Windows · x64
SanFrame_6.30.6_x64-setup.exe
Use this to check that your download matches the published file.
dab95e1e10619d458550fce0bd5c75666b103886361055da123711059b63afc3macOS · 13.4+
SanFrame_6.30.6_universal.dmg
Use this to check that your download matches the published file.
aac1ef4273e86db09562387eadfe8ca1600f80a967a43a7ff427d2f6768d6790Android 10+ · APK · 2.85 MiB
Bring your phone into your workflow. Share text, images and files with your computer over the same LAN.
Invite your phone from the desktop app and confirm the pairing codes to connect.
SanFrame传输-5.0.0-release.apk
Use this to check that your download matches the published file.
da27f37664cef4bd1d2cbfba88908d493f56c9c0a57cbcf657b1878900efaadc