SELF OS Personal Intelligence

Home repository: github.com/selflabs/SELF-OS — canonical source for issues, PRs, and releases of SELF OS Personal Intelligence.

An open-source personal intelligence shell for building, running, and customizing AI agents, apps, and device-native workflows.

SELF OS Personal Intelligence is the open developer layer of SELF OS — a personal intelligence environment where people can install agents, run local AI tools, connect apps, and build custom automations around their life.

The open-source edition gives developers and power users the foundation to shape their own agent-powered personal intelligence setup on-device.

The full SELF OS experience adds managed agents, Harmony Mesh access, ESSENCE rewards, Aura data monetization, phone services, cloud sync, and production-grade security.

Tagline: Build your own personal intelligence system — or connect to the full SELF OS ecosystem when you are ready.

Platforms: SELF OS Personal Intelligence is a mobile layer — meant to run on the devices people carry every day. Android is available now in this repository (phone and tablet). Apple (iOS) is coming soon; the same product vision applies across platforms once the iOS edition ships.

Model providers: You are not locked into one AI vendor. As you extend agents and apps, you can integrate different model providers — for example Resonatia (quantized models through a REST chat API; authentication and request shape are documented there), OpenRouter, OpenAI, Anthropic (Claude), Google (Gemini), and other compatible APIs or on-device runtimes. Community Edition ships deterministic example agents only; wiring a provider means adding your own client code, credentials, and privacy review. See docs/LOCAL_MODELS.md for local vs cloud notes.

---

1. What is SELF OS Personal Intelligence?

Community Edition is a standalone Android app and SDK workspace today: a reference mobile personal intelligence shell that shows how the open layer fits together. (An iOS edition is planned next.)

• A shell UI for your personal intelligence workspace (home, agents, apps, mesh preview, wallet preview, settings).
• On-device agent execution via AgentRuntime — core to a local-first personal intelligence stack.
• SELF Apps in a PluginRegistry, so intelligence capabilities can grow like mini-programs around the user.
• Mock Harmony Mesh, wallet, app store, and Aura-style consent previews — safe stand-ins while you design real personal intelligence flows.

Live SELF OS backends are intentionally out of scope here. See docs/OPEN_SOURCE_BOUNDARY.md.

2. What can you build for personal intelligence?

• Agents that implement the Agent interface — specialized reasoning or helpers inside your intelligence shell (examples/agents/). Agents can call your choice of model backends (e.g. Resonatia, OpenRouter, OpenAI, Claude, Gemini, or local models) once you add the integration.
• SELF Apps via SelfApp — surfaces and tools that extend what the user can do from one personal intelligence home (examples/apps/).
• Custom mocks under mock/ so prototypes keep the same contracts without touching private infrastructure.

3. SELF OS Personal Intelligence vs SELF OS Core (full platform)

This repo: open personal intelligence edition	SELF OS Core (separate): full managed intelligence & platform
Local shell UI & SDKs for your own PI setup	Live Harmony Mesh routing
Mock economics & tasks (preview only)	ESSENCE earning & settlement
Educational Aura toggles (consent UX)	Aura monetization programs
No telecom provisioning	Telnyx / phone services
No managed cloud agents	Orchestrated cloud workloads

4. Features (personal intelligence experience)

• Mobile-native Android shell — Material 3 / Jetpack Compose, six main destinations plus agent chat, tuned for daily personal intelligence on handhelds and tablets (iOS version coming soon).
• Theme controls (system / light / dark) so the shell feels like your environment.
• Agent chat wired to AgentRuntime.sendRequest — conversational layer over local intelligence; swap in real LLMs via the providers you configure (OpenAI, Claude, Gemini, OpenRouter, Resonatia, etc.).
• Mock app catalog — practice how apps plug into a personal intelligence launcher (in-memory).
• Safety-first layout for an open PI project: no google-services.json, no keystores committed in-tree.

5. Architecture of the open intelligence layer

The codebase is modular so you can evolve personal intelligence without a monolith. See docs/ARCHITECTURE.md.

• apps/self-shell — the Android personal intelligence application (Gradle module name unchanged).
• sdk/agent-sdk, sdk/app-sdk — contracts for agents and apps that live in the user’s PI stack.
• mock/* — local-only stand-ins for mesh, wallet, store, and data-consent previews.

6. Quick Start (Android — run the app on a device or emulator)

This Quick Start is for Android; iOS setup will be documented when the Apple release is available.

Clone from GitHub (or open the public-self-shell folder if it lives inside your monorepo):

git clone https://github.com/selflabs/SELF-OS.git
cd SELF-OS

Requirements: JDK 17, Android SDK (compileSdk 34), and a physical Android device or emulator.

./gradlew :self-shell:assembleDebug

On Windows:

.\gradlew.bat :self-shell:assembleDebug

Open the project in Android Studio to iterate on your personal intelligence UI and agents.

Monorepo maintainers: see docs/GITHUB.md for pushing this tree to selflabs/SELF-OS.

7. Creating an agent (intelligence primitive)

Agents are how the shell does work on behalf of the user. Read docs/AGENT_DEVELOPMENT.md and study:

• examples/agents/hello-agent/HelloAgent.kt
• examples/agents/task-planner-agent/TaskPlannerAgent.kt
• examples/agents/local-notes-agent/LocalNotesAgent.kt

8. Creating a SELF App (intelligence surface)

SELF Apps are installable surfaces in the same personal intelligence workspace. Read docs/PLUGIN_DEVELOPMENT.md and study:

• examples/apps/hello-self-app/HelloSelfApp.kt
• examples/apps/personal-dashboard/PersonalDashboardApp.kt

9. Mock services (safe personal intelligence sandbox)

By default there is no live link to SELF OS Core. Mocks under mock/ give deterministic previews so you can design personal intelligence flows without keys or production endpoints. Details: docs/MOCK_SERVICES.md.

10. Open-source boundary

What stays in the open personal intelligence edition vs what remains in SELF OS Core: docs/OPEN_SOURCE_BOUNDARY.md.

11. Roadmap

Where the open personal intelligence layer may go next: docs/ROADMAP.md.

12. Security

Protecting users and forks of this personal intelligence codebase: SECURITY.md. Before release, run:

• scripts/scan-public-safety.sh or scripts/scan-public-safety.ps1

13. Contributing

Help improve SELF OS Personal Intelligence: CONTRIBUTING.md and CODE_OF_CONDUCT.md.

14. License

SELF OS Personal Intelligence is licensed under the GNU Affero General Public License v3.0 — see LICENSE. See also NOTICE.

---

Preview disclaimer (also shown in-app): Preview only. No real earnings are generated in SELF OS Personal Intelligence (Community Edition).