Persistent Sage is a privacy-oriented desktop AI companion: multi-thread chat, long-term Memory Anchor storage, optional agent tools, customizable companion personalities, Pulse scheduled check-ins, vision image attachments, and Coding mode — a repo-scoped development workspace with editor, terminal, and git tools—all in a local-first Tauri 2 application.

Repository: github.com/g00siferdev-py/persistent-sage
Status: Persistent Sage 3.0.0 — GitHub tag v3.0.0; Microsoft Store 3.0. Feedback welcome via GitHub Issues

Version 3.0.0 — Moltbook, Productivity mode, Favorites, Share, PDF tools, workspace vision, experimental subagents.

Maintainers: publish installers → docs/PUBLISH.md; MSIX → docs/MICROSOFT-STORE.md; CI → docs/BUILD-CI.md.

Settings → Companion includes Import OpenClaw markdown…, but the most reliable way to move a mature OpenClaw personality into Persistent Sage today is:

1. Copy IDENTITY.md, SOUL.md, JOURNAL.md, USER.md, and MEMORY.md into Persistent Sage’s workspace/ folder (enable workspace tools).
2. Prompt the companion to read those files and update personality.json via personality self-edit (enable in Settings → Tools).
3. Remove the .md files from workspace/ when done.

We are still improving one-click migration; see docs/USER-GUIDE.md § Migrating from OpenClaw and CHANGELOG.md for the exact prompt and details.

After you build and run Persistent Sage, nothing is stored on a Persistent Sage-operated cloud. Messages go only to the LLM provider you configure (and optional tool URLs if you enable agent tools). See PRIVACY.md and docs/DATA-AND-PRIVACY.md for the full picture.

• 3.0 — Moltbook, Productivity mode (Google widgets / Email Agent), Favorites, Share, agent PDF tools, workspace vision (attach from workspace + workspace_view_image), experimental subagents, Settings tabs (General / Provider / Tools).
• Coding mode (v2 / 2.1+) — Git repos, editor, terminal, coding agent, playground, notepad, Agent Action Stream. See docs/CODING-MODE.md.
• UX (2.1+) — Message timestamps, Help menu, token counter, light/dark theme, cache manager, abort turn, unified Companion↔Coding context; 3.0 restores fenced code blocks and reliable post-tool replies.
• Memory Anchor — SQLite conversations, messages, anchors, projects, and preferences; hybrid FTS + keyword (+ optional semantic) recall and startup briefings.
• Companion profiles — Multiple personalities with live system-prompt preview; Persistent Sage JSON and OpenClaw markdown import; optional agent self-edit of personality.json.
• Providers — OpenAI, Google Gemini, xAI Grok, OpenRouter, Ollama (local), Ollama Cloud, Anthropic, or offline placeholder.
• Agent tools (opt-in) — Web search, URL fetch, headless fetch_browser, HTTPS http_request, sandboxed workspace files & images, PDF read/create, optional database query, Moltbook, experimental subagents.
• Pulse — Timer-driven check-ins that run as normal chat turns in your selected sidebar thread.
• Vision — Attach images (computer, workspace, or webcam); multimodal payloads for supported models.
• In-app updates — Settings → General → Updates for both Store and GitHub installs (Store APIs vs Tauri updater).
• Send feedback — Settings buttons open prefilled GitHub Issues without attaching private chats or logs.
• Optional donations — PayPal / Cash App links in footer and onboarding (no feature unlock).
• Portable layouts — PERSISTENT_SAGE_DATA_DIR and PERSISTENT_SAGE_PORTABLE (legacy NOVA_* also works) for custom or USB data locations.

git clone https://github.com/g00siferdev-py/persistent-sage.git
cd persistent-sage
npm install
npm run tauri dev

First launch creates local data under your OS app directory (or PERSISTENT_SAGE_DATA_DIR if set). Configure Settings → Provider, then start a chat.

New to the stack? Follow the step-by-step guide in docs/INSTALL.md.

export PERSISTENT_SAGE_DATA_DIR="$HOME/PersistentSageData"
mkdir -p "$PERSISTENT_SAGE_DATA_DIR"
npm run tauri dev

Persistent Sage was developed in Cursor with heavy AI pair-programming. Throughout the project, coding work was driven by selecting Codex and GPT-5.6 as the models inside Cursor.

Those two models played a huge part in writing Persistent Sage: implementing the Rust backend and React UI, Memory Anchor / SQLite recall, provider integrations, agent tools (including PDF), Coding mode, packaging, and docs. Direction, product decisions, review, and release ownership stayed with the maintainer; Codex and GPT-5.6 were the primary coding partners in the editor.

More: docs/INSTALL.md § Troubleshooting

Persistent Sage 3.0.0 (beta, tag v3.0.0-beta.1) ships Moltbook, Favorites, Share/copy enhancements, agent PDF tools, personality update improvements, webcam capture, and a refactored Settings UI — on top of Companion + Coding from 2.x. GitHub beta installers are available now; Microsoft Store 3.0 is planned for Aug 3. See PERSISTENT-SAGE-STATUS.md, CHANGELOG.md, and docs/releases/v3.0.0.md.

Maintainer: g00siferdev-py

MIT License — Copyright (c) 2026 g00siferdev-py