SScreenSearchScreen Memory
Download
Local AI Screen Memory

A local second brain built from your Windows screen history.

Turn changed screens captured within your settings into a searchable layer beside your notes. Use OCR, hybrid search, Smart Answers, and Recall reports to find past text and inspect the source frames behind an answer.

Build Your Second Brain Learn More
Windows 10/11 · Local-first · No account
Bundled llama.cpp AI · no remote inference provider
01 — A Second Brain From Your Real Work

A Second Brain From Your Real Work

Notes hold what you deliberately save and shape. ScreenSearch complements them by making captured windows, documents, terminals, and pages searchable when you did not stop to take a note.

Notes and deliberate knowledge

  • Capture the ideas you intentionally choose to keep
  • Shape durable knowledge with your own structure, links, and commentary
  • Support rich search and retrieval inside the note system you already use
  • Remain the right home for conclusions, plans, and polished reference material

ScreenSearch alongside your notes

  • Setting-controlled capture records changed screens while respecting pauses and exclusions
  • UIA text plus OCR fallback reads apps, images, PDFs, browsers, and terminals
  • Hybrid semantic search finds conceptually related content, not just keywords
  • Source-linked results let you inspect the captured frame, app, and time behind an answer

What Makes It a “Second Brain”?

Remembers Useful Context
Attention-first capture keeps the searchable history focused on work, not repetitive chrome.
Connects Ideas
Vector search and keyword search find related concepts automatically.
Answers Questions
Ask questions, get grounded answers, and inspect the source frames.
Local-First Control
Capture, OCR, indexes, storage, and AI inference stay on your Windows PC.
02 — Local AI, Practical Recall

Local AI, Practical Recall

Current screen-search capabilities and opt-in local extension surfaces.

  1. T.01

    UIA text + OCR fallback

    ScreenSearch reads accessible app text first, then falls back to OCR when content only exists inside pixels. That keeps browsers, PDFs, terminals, images, and older apps searchable.

    Windows UIA -> OCR fallback

  2. T.02

    Attention-first enrichment

    Surface reduction and smart enrichment throttling cut down repetitive UI chrome before summaries and search indexes are built.

    Focused capture · smarter throttle

  3. T.03

    Hybrid Search

    Keyword search and vector search are fused with Reciprocal Rank Fusion, so exact phrases and nearby ideas can meet in the same results.

    FTS5 + vectors -> RRF

  4. T.04

    Smart Answers + Recall reports

    ScreenSearch retrieves matching screens, then your chosen model drafts answers and reports linked to source frames so you can check the result.

    Search -> Retrieve Context -> Answer

  5. T.05

    Bundled Local AI

    Answer generation runs through the bundled llama.cpp sidecar on a loopback address. Current releases do not expose a remote inference-provider setting.

    llama.cpp sidecar · 127.0.0.1

  6. T.06

    Flow overlay + local API

    Use the Flow overlay hotkey to search or ask without switching apps. Opt-in localhost API, export, and MCP surfaces support trusted local workflows.

    Ctrl+Alt+Space · API/export/MCP

03 — AI-Powered Use Cases

AI-Powered Use Cases

Search captured work, draft source-linked answers, and verify the result against the original frames.

“Summarize what I worked on this week”

Ask a natural-language question. ScreenSearch retrieves matching captured screens, then your chosen model drafts a summary linked to source frames. Review those sources before using the result for a standup, report, or reflection.

You: "What did I work on this week?"AI:  "Monday: Debugged authentication. Tuesday: Built dashboard..."

“Find all references to database migrations”

Hybrid search can connect an idea with related wording. Search “database migrations” to surface captured screens that mention terms such as “schema changes,” “SQL ALTER TABLE,” or “DB versioning.”

Semantic matches: "schema evolution" "DB versioning" "migration scripts"
04 — Second Brain FAQ

A searchable layer beside your notes

Review how local screen search works, or read the privacy details for capture, retention, exclusions, local AI, and network use.

Does ScreenSearch replace my notes app?
No. Notes remain useful for deliberate ideas, decisions, and durable reference material. ScreenSearch complements them by making captured on-screen context searchable when you did not save it manually.
What becomes part of the second brain?
Only changed screens recorded within your capture settings. You can pause capture, exclude apps or windows, and apply retention controls to the local history.
Can I verify an AI answer?
Smart Answers and Recall reports are built from retrieved screen history and link back to source frames. The model can still be incomplete or mistaken, so inspect the source before relying on a result.
Does AI processing stay on my PC?
Yes. Capture, OCR, search indexes, storage, retrieved context, and answer generation stay local. AI requests run through the bundled llama.cpp sidecar on the local machine; current releases do not expose a remote inference-provider setting.

Build Your AI Second Brain Today

Add searchable screen history beside your notes, then verify retrieved context against the captured source frames.

Download for Windows View on GitHub
Hybrid search Local LLMs No account