ClipBrain dashboard showing saved captures, tags, summaries, and source types
Local-first memory for AI agents

ClipBrain

Clip anything into agent-ready local memory.

Web pages, X posts, Kindle highlights, YouTube transcripts, Gmail threads, and PDFs become searchable context for Codex, Claude Code, Claude Desktop, Cursor, and any MCP client.

  • No ClipBrain cloud
  • MCP-ready
  • Runs on 127.0.0.1

Capture layer

Your reading life becomes queryable infrastructure.

ClipBrain is two parts: a Chrome extension for capture and a local server that stores, compiles, and exposes your knowledge over MCP. The extension is the door; the local memory layer is the product.

One shortcut, every source.

Press Cmd+Shift+S or use the toolbar button. ClipBrain extracts the page, sends it to your local server, enriches it if you configured AI processing, and keeps the raw source searchable.

  • Articles are cleaned with Readability.
  • Kindle imports walk every book and save highlights plus notes.
  • Context packs return cited sources your agent can use immediately.
Chrome extension popup capturing an X post into ClipBrain

Web

Clean article text, title, URL, tags, summaries, and source chunks.

Kindle

Highlights, notes, book metadata, and chapter-level reading evidence.

YouTube

Transcripts with timestamps, ready for search and summarization.

PDFs

Drag, drop, extract, index, and ask your agent about the file later.

Gmail

Optional permission for threads you explicitly choose to capture.

Kindle notebook page with imported highlights and the ClipBrain import control

Kindle highlights

Your books stop living in a separate silo.

Open your Kindle notebook, click Import to ClipBrain, and the extension walks the library for you. Highlights become searchable memory with book titles, notes, and citation-ready source text.

Agent workflow

The magic is not storage. It is context on demand.

ClipBrain uses GBrain under the hood for local knowledge storage and retrieval, then adds capture workflows and a ClipBrain-specific context pack for compact, cited agent handoffs.

01

Capture

Save the active page, thread, transcript, PDF, or Kindle highlight set into your local ClipBrain server.

02

Compile

ClipBrain preserves raw text, creates summaries and tags when configured, and keeps long sources searchable in chunks.

03

Ask

Your MCP client can query your corpus and receive cited context packs with sources, snippets, claims, quotes, and actions.

Install

Open source, local server required.

The Chrome extension is the capture half. It needs the ClipBrain local server running on your machine at 127.0.0.1:19285.

git clone https://github.com/agentpilled/ClipBrain
cd ClipBrain
./setup.sh
bun run doctor

Extension

Install the published extension from the Chrome Web Store. For local development, load the repo folder unpacked from chrome://extensions.

MCP clients

Setup configures the local memory tools for compatible AI clients, including the ClipBrain context_pack handoff.

Privacy boundary

No hosted ClipBrain backend.

Your captures go from Chrome to your local server. If you configure optional AI enrichment with your own OpenAI key, your local server may send captured text to OpenAI for summaries and tags. ClipBrain itself does not run a cloud corpus.

StorageYour machine
Extension endpoint127.0.0.1
Cloud accountNot required
AI enrichmentOptional, user-configured

Give your agents the context you already earned by reading.

ClipBrain turns the internet, your books, and your documents into local memory your AI can actually cite.