Comparison, AI PC rationale, tradeoffs, costs, and deployment paths for the local compounding engine between your brain and your AI clients.
Augur is a local harness layer, not an LLM wrapper, and it does not require an Augur API key. Claude, Codex, Gemini, Cursor, and Ollama remain the native thinking engines you already use. Augur is the local MCP and compounding layer that lets those engines operate the same brain and write useful outcomes back into durable files.
Your existing subscriptions and local models provide reasoning. Augur gives them one persistent brain to work against.
The adapter is exposed locally: memory, skills, files, workflows, approvals, and tools stay inspectable instead of disappearing into a cloud wrapper.
Switch models or remove Augur without losing the notes, wiki, memory, prompts, skills, workflows, and source cards that make the system yours.
Most people start in Operation mode. Some grow into Dev mode. Both are valid — the system meets you where you are.
Not coding — but how your system works. YAML files. Markdown. How skills connect. You build trust because you can see exactly what your AI knows.
Augur keeps you in decisions that matter. More time, but you remain the architect.
When something breaks, you fix it. Uncomfortable — and how you grow.
People sync files to their PC even when they have cloud storage — because important files only on the cloud feels risky. Your second brain is higher stakes than files.
The persistent home for cross-session memory, skills, workflows, and business logic. No cloud database, no sync latency — your second brain lives on flash storage you control.
The orchestrator and integration manager. Connects CLIs to AI providers, runs the MCP server, manages the daemon, spawns processes. The CPU makes the PC the integration hub.
Full offline reasoning via quantized local models. When you need capability without internet, the GPU keeps Augur fully functional in airplane mode.
Real-time, privacy-critical inference: speech-to-text via Whisper, OCR via Tesseract. These tasks need low latency and must never leave the device.
| Aspect | Cloud-only AI | Augur |
|---|---|---|
| Your data & memory | Stored on provider's servers | On your disk — always |
| Workflows & skills | Locked to one platform | Live repo skills and private vault drafts (open SKILL.md format) |
| LLM provider | Single vendor lock-in | Bring your own — Claude, GPT, Ollama, any |
| Switch providers | Lose everything | Keep all data, swap the AI |
| Privacy | Provider's terms of service | Your files never leave your machine |
| Customization | Limited to provider UI | Live repo skills, private vault drafts, community packs, full dashboard |
| Overnight ops | Costs money per API call | Auto-commands run free on idle hardware |
| Offline mode | No internet = no AI | Full airplane mode with local LLM via Ollama |
Augur is the persistent layer. The AI is replaceable.
Start on the homepage for early access, or open the public Augur OS repository for the philosophy, architecture, and implementation details.