Your Second Brain.
Locally Owned. Blazing Fast. Built for Humans & AI.
A local-first, encrypted knowledge sanctuary designed for deep human reflection and autonomous AI agent memory. Standard plaintext Markdown on your disk, instant sub-millisecond search, under 50MB RAM (<25MB local-only), and zero cloud lock-in.
Definition: Sphene is the reference implementation of an Agent-Native Knowledge Kernel—a local headless markdown knowledge base engineered with native Model Context Protocol (MCP), local-first offline PWA, and an AST-level Differential Timeline for zero-token-waste agent interactions under 50MB RAM (<25MB for single PC/local AI agents).
Choose Your Starting Point
Pick the experience that fits your workflow today. Both deliver the exact same beautiful interface, and you can connect them at any time.
Sphene Local (Standalone PWA)
For thinkers, writers, and researchers who want an instant, private note-taking space. Runs 100% locally in your browser sandbox. Your notes never touch our servers — stored entirely on your device with zero cloud tracking.
Sphene Core (Server)
For builders, homelabs, and power users who want total physical data sovereignty and native Model Context Protocol (MCP) for AI agents.
curl -fsSL https://raw.githubusercontent.com/Sphene-app/scripts/main/install.sh | bash
Engineered from First Principles
Three uncompromised design tenets delivering whisper-quiet efficiency, sub-millisecond retrieval, and verifiable data sovereignty.
Native Compiled Efficiency
Engineered as a single, lightweight native binary with zero runtime dependencies. Operates as a local-only daemon at <25MB RAM (~10–14MB idle) for single PC and local AI agent setups. When mesh networking is enabled, the embedded WireGuard relay remains strictly <50MB RAM total.
Sub-3ms End-to-End Relational Retrieval
Powered by an embedded SQLite relational core with native FTS5 full-text indexing. Document searches resolve in 2.20 milliseconds end-to-end via HTTP/MCP API (~180µs raw internal FTS5 index lookup in memory), maintaining instant responsiveness across vaults with thousands of notes.
Standard Markdown & Total Privacy
Every note lives directly on your computer as standard plaintext Markdown (.md). For sensitive journals, finances, or credentials, lock your vault with authenticated client-side AES-256-GCM zero-knowledge encryption—complete with an independent recovery runbook so you are never locked out.
Engineering for the Agent Era
Deep dive into native Stdio MCP over container proxies, AST-level section manipulation, and the Differential Timeline with Human Veto.
Verifiable Systems Performance
Hard metrics measured across 1,000 markdown notes: 2.20ms end-to-end API search (180µs raw FTS5 index), <1.5ms MCP tool execution, and 31MB–40MB RAM footprint at rest.
Designed from the Ground Up for Autonomous AI
Connecting AI agents shouldn't mean letting bots silently corrupt your knowledge. Sphene gives agents instant memory recall while keeping you in total control.
Native Agent Gateway (MCP & REST/SSE)
Connect Claude Desktop, Cursor, or Hermes Agent in seconds via official Model Context Protocol (MCP) over stdio, local sockets, and real-time change streams.
The Differential Timeline ("Human Veto")
Never fear AI hallucinating or overwriting your notes. Agent edits are cleanly staged at the block level for your visual review with 1-click accept or rollback.
Hardware-Grade Vault Partitions
Strict architectural boundaries: Workspace/ (collaborative human & agent), Reference/ (write-locked facts/books), Private/ (encrypted, 403 Forbidden to AI), and Drawing/.
Drawing Studio & Visual Canvas
Infinite visual canvas for architecture diagrams, flowcharts, and sketches. Compiles directly to native Mermaid syntax and vector Markdown.
Private On-Device AI Assistant
Instant note synthesis, contextual Q&A, and semantic backlinks computed locally on your device with complete privacy — zero external API keys or subscriptions required.
Interactive 3D Knowledge Graph
Visual force-directed knowledge graph rendered directly from the engine. Smooth traversal across [[Wikilinks]], tags, and backlinks without dragging down browser performance.
Local-First Standalone PWA & Sovereign Cloud Sync
Your second brain isn't trapped on your desktop. Experience native mobile speed, seamless client-to-cloud backup, and end-to-end encrypted mesh relays with zero proprietary gatekeepers.
Native Modern PWA (iOS & Android)
Install Sphene as a standalone desktop or mobile app on iOS, Android, macOS, and Windows. Powered by client-side IndexedDB caching, full offline capabilities, and zero app store middlemen.
Sovereign Cloud Sync (GDrive & Dropbox)
Sync your standard Markdown notes across all your devices using your own Google Drive or Dropbox account. Direct client-to-cloud synchronization means your data never routes through third-party servers.
Zero-Config Tailscale WireGuard Mesh
Pair your phone, laptop, and home server seamlessly with embedded Tailscale mesh networking. End-to-end WireGuard encrypted peer-to-peer tunnels with automated HTTPS, zero port forwarding, and strict <50MB total footprint.
Collaborate with Absolute Peace of Mind
Sphene Teams allows co-workers, engineering squads, research partners, and families to share notes securely. Backed by high-durability S3 storage, your shared knowledge stays protected with device-level zero-knowledge encryption.
Device-Level Encryption
Notes are encrypted directly on your device before they ever leave your hardware. No one—not even Sphene engineers or cloud storage providers—can read what your team writes. Only teammates with your private passphrase can decrypt your words.
S3-Backed Cloud Durability
Sync is backed by high-durability S3 storage with multi-region redundancy. If you close your laptop or switch to your phone, your encrypted notes sync smoothly across all connected teammate devices automatically.
Instant 1-Click QR Pairing
Invite teammates or pair your phone with a simple camera scan. No complicated setup, no passwords to email, and no third-party tracking accounts required. Point, scan, and start collaborating in seconds.
Test team collaboration with code freeSpheneSync. Plenty of room for shared sprints, docs, and knowledge bases with zero cost.
Engineered for Sovereign Thought
Zero mockups or placeholder graphics. Authentic, high-resolution captures from a live research vault running on the Sphene sovereign kernel.
Autonomous Agent Architecture & Editor
Clean distraction-free typography with inline Mermaid vector diagrams, collaborative Human-Veto staging, and hardware partition indicators.
Featherweight On-Device Note Q&A & Summarizer
Instant conversational Q&A and key point extraction running privately on your device. Lightweight statistical analysis executes in milliseconds with zero cloud roundtrips and zero battery drain.
Force-Directed Relational Graph
Dynamic cluster visualization color-coded by Aegis partitions. Trace conceptual lineages across literature, agent specs, and private vaults smoothly.
Antracite Theme: Mathematical Formulations
Subdued deep anthracite-slate styling tailored for technical researchers, vector embedding mathematics, and sub-0.20ms benchmark audits.
Cyber Velvet Theme: Homelab DevOps
High-energy contrast for systems engineering, Traefik topologies, and immutable ZFS backup checklists.
Native Vector Drawing Studio
Draw architectural flowcharts and whiteboard concepts directly inside your vault. Sticky connection points stay anchored when boxes move.
The Complete Toolkit, Built-In. Zero Setup Fatigue.
Skip the endless plugin hunting, fragile dependency chains, and broken updates. Sphene delivers a cohesive, verified suite of capabilities directly out of the box — visual canvas, on-device AI synthesis, relational search, LaTeX math, and 1-click desktop vault importers.
🎨 Visual & Canvas
Infinite Drawing Studio, Mermaid SVG flowcharts, and Interactive Image Lightbox gallery.
🧠 Sovereign AI
Private On-Device Assistant, Cosine Semantic Backlinks, and Smart Inbox Classifier.
📥 Knowledge & Importers
Bookify Article Sanitizer & Chapter Splitter, Kindle Highlights Bridge, Google Keep, and Google Drive sync.
🛡️ Security & Backup
1-Click Full-Vault Lock, Signed Ed25519 Git Commits, and Sovereign Trash Bin.
📖 Reading & Typography
KaTeX Fast Formula Renderer, Floating Dynamic Table of Contents, Zen Focus Mode, and 5 Curated Themes.
⚡ Productivity & Tasks
Interactive Markdown Table Builder, 365-Day Note Streak Heatmap, and Vault-Wide Task Matrix.
Architectural Dimension Matrix
Comparing a sovereign compiled substrate with traditional desktop PKMs and centralized cloud platforms.
| Architectural Dimension | Heavy Desktop Web Runtimes (Electron / Webview) | Centralized Cloud Platforms (SaaS) | Sphene Sovereign Hub & PWA |
|---|---|---|---|
| Primary Sweet Spot | ✓ 1,000+ community extensions, themes & visual canvas plugins | ✓ Multi-table relational databases, rollups, sprint boards & real-time team collaboration | ✓ Sub-3ms search, 31MB–40MB RAM footprint, native AI agent MCP & zero cloud lock-in |
| Physical Data Custody | ✓ 100% Plaintext Markdown (.md) on local filesystem | 0% local files (proprietary database blocks stored in vendor cloud) | ✓ 100% Plaintext Markdown (.md) on local NVMe/SSD + SQLite WAL |
| Physical RAM at Rest (1,000 notes) | 797.73 MB – 805.57 MB (6 OS background processes: Main, GPU, Renderer, Helpers) | 350 MB – 650 MB (Per active browser workspace tab) | ✓ 30.76 MB – 40.46 MB (1 OS process, 19.9x – 26.2x lighter) |
| Search Index Latency (1,000 notes) | 86.30 ms (Filesystem directory scan / regex) | 180 ms – 450 ms (Remote cloud API roundtrip over WAN) | ✓ 2.20 ms (SQLite FTS5 WAL index, 39.2x faster) |
| Single Note Read / Write Latency | 18.40 ms read / 24.60 ms write (Electron IPC hop + cache) | 150 ms – 500 ms (HTTPS WAN request) | ✓ 0.20 ms read / 1.20 ms write (WAL memory transaction) |
| Autonomous Agent Interface | Third-party REST wrappers (120ms–380ms subprocess / curl) | Hosted webhooks (Proprietary cloud API fees) | ✓ Native MCP stdio / SSE (Direct JSON-RPC, <1.5ms tool roundtrip) |
| AI Concurrent Write Safety | 0% collision safety (Blind silent disk overwrite; in-flight human edits lost) | Server-side revision log (Cloud-dependent version history) | ✓ 100% collision-proof (Differential Timeline: Human Veto diff staging) |
| Agent Context Overhead (1,000 notes) | 100% whole-file dump (52,000+ tokens transmitted to LLM) | Opaque token pricing (Server-side hosted rate limits) | ✓ 56.5% – 71.2% token savings (Targeted section-level AST isolation) |
| Structured Databases & Rollups | Third-party community query scripts | ✓ Native multi-table rollups, formulas & kanban | YAML frontmatter + embedded sandboxed query engine |
| Client-Side Cryptography | 0-bit unencrypted plaintext files on disk | Proprietary vendor-managed cloud keys | ✓ Standard .md + Client-Side AES-256-GCM Vault Lock |
- Choose Centralized Cloud Platforms if you need complex multi-table relational databases, accounting rollups, project sprint boards, and collaborative multi-user team workspaces.
- Choose Heavy Desktop Web Runtimes if your workflow depends on thousands of community-built desktop themes, canvas plugins, and third-party UI extensions.
- Choose Sphene Sovereign Hub if you prioritize raw local speed (2.20ms indexed search), whisper-quiet memory (31MB–40MB RAM), native AI agent memory (MCP for Claude Desktop, Cursor, Hermes) with staged diff review ("Human Veto"), client-side AES-256-GCM authenticated encryption, and zero cloud vendor lock-in. Your notes remain 100% sovereign Markdown (.md) files that can be opened in any text editor at any time.
Empirical Benchmarks →
2.20ms SQLite FTS5 search latency, 31MB–40MB RAM footprint, and reproducible test harness.
Sovereign Architecture →
Compiled Go substrate, embedded SQLite WAL indexing, capability sandboxing, and PWA synchronization.
Aegis Security & Partitions →
Client-side AES-256-GCM vault encryption, Ed25519 signatures, and 4-zone vault isolation boundaries.
Take Full Ownership of Your Second Brain
Deploy locally in 60 seconds. Free and open-format forever. Expand to managed cloud sync when you are ready.
