
notabene
Review Markdown docs by commenting on the rendered page, then have an AI agent apply each note to the source and journal it, with git diffs to approve. No SaaS.
READ →

Free up disk space on a Mac from the terminal: caches, apps and their leftovers, stale node_modules, with a --dry-run preview before anything is deleted.

Keep invoices and contracts on your own server and find them by their text: OCR, email or watched-folder intake, one Docker container running on SQLite.

Follow a busy story without rereading every syndicated copy: a self-hosted RSS reader that groups articles into events using local or remote AI models.

Word, Excel, PowerPoint and text PDFs to Markdown for an LLM, with no Python and no ML model: a Rust library, a Node CLI, a WebAssembly build. No local OCR.

Ask an Obsidian vault questions and get cited answers: a Python graph RAG on Ollama with an MCP server and a 3D view, though its default LLM runs in the cloud.

Compile documents once into a Markdown wiki an LLM keeps current, rather than re-reading them for every question. GPL-3.0 desktop app with graph and MCP server.

Turn a website or a small Node.js, PHP, Python or Go server into a signed Android APK, built entirely on the phone, without a PC or a remote build queue.

Describing a spot in sentences, quoting a passage to remove doubt: a prompt can ask, but it can't point. How notabene turns page comments into agent tasks.

Run branching, multilingual surveys on your own PHP server under the GPL, or in the vendor's cloud, which is priced by the number of responses you collect.

Typeform-style forms on your own server with no cap on responses. The free self-hosted edition allows two accounts per instance; a paid license lifts the limit.