Recall
A self-hosted AI notebook that gives any assistant — Claude, ChatGPT, Gemini — persistent memory of your drafts, research, and ideas, stored as plain Markdown on your own machine and never synced to the cloud.
writers, journalists, and content creators working on long-term projects who use AI daily but won't send unpublished work to third-party servers
- Markdown-based vault of every conversation and draft, version-controlled
- One-click connection to Claude Code, Cursor, ChatGPT, or web chat
- Hybrid search across everything you've ever written with the AI
- Local vector index and export — no cloud sync, no telemetry
Writers increasingly lean on AI for drafts but worry about unpublished manuscripts living on cloud servers; self-hostable tools like AsterMem (86 stars) prove the niche is real.
Multiple open-source self-hosted AI-memory projects (AsterMem, personal-rag, claude-obsidian) and several recent Medium/dev.to tutorials show real, active interest from writers and devs in persistent Markdown-backed AI memory.GitHub - pantgr/personal-rag: Self-hosted Personal RAG ↗How I gave my AI assistant persistent memory using nothing but markdown files ↗
Niche is crowded: AsterMem, personal-rag, Open Notebook (self-hosted NotebookLM clone), claude-obsidian, and Obsidian plugins (Smart Connections, Copilot) all already target writers wanting private AI memory on Markdown.Open Notebook: The Open Source Alternative to NotebookLM ↗claude-obsidian: Self-Organizing AI Second Brain for Obsidian - GitHub ↗Obsidian + AI in 2025: Smart Connections vs. Copilot Plugin vs. Claude ↗
Writers pay Obsidian ~$4-8/mo for Sync/Publish around local Markdown, but competing self-hosted AI-memory tools are open-source/free, making paid conversion difficult; no clear paid product in this exact niche surfaced.Obsidian Pricing Breakdown: Prices, Plans, Pros, and Cons ↗AI Research Tools & Data Privacy: What Happens to Your Unpublished Work ↗
Local-first, privacy-first, and self-hosted AI tool trends are accelerating into 2026 with multiple new launches and growing media coverage, suggesting a multi-year tailwind rather than a fad.Building a Self Improving AI Assistant from 10 Markdown Files ↗How I Gave My AI Persistent Memory: From Markdown Hacks to MCP ↗
Existing OSS projects (AsterMem, personal-rag, Open Notebook) already implement the core stack: local Markdown + vector/RAG + Docker; MCP integrations with Claude/ChatGPT/Gemini are documented and reusable.GitHub - Asterove/AsterMem ↗Open Notebook: The Self-Hosted AI Research Tool ↗