Both platforms run AI agents on PCs. The difference is what your hardware team and customers actually get.
Data stays on the PC. No cloud dependency for core functions. Your users' data is their own.
IrisGo observes how users work and builds context over time. Agents get smarter the longer your hardware runs.
Intelligent routing between on-device and cloud. Your hardware specs become a selling point, not a cost line.
What matters to OEM partners and their customers
|
IrisGo
Managed Platform
|
OpenClaw
Open Framework
|
|
|---|---|---|
| Architecture | ||
| Processing Model |
Hybrid compute, on-device first Intelligently routes tasks between local models and cloud APIs. Sensitive data stays on-device; heavy tasks go to cloud when needed. |
Cloud-dependent Requires API calls to cloud LLMs for most operations. Local mode limited. |
| OEM Integration |
Turnkey pre-install White-glove onboarding. IrisGo brand stays visible — like Intel Inside for AI. Co-branded experience, our engine underneath. |
DIY assembly OEM must build UI, user flows, update pipeline, and support stack from scratch. |
| Model Support |
Multi-model orchestration Intelligent routing across local NPU models, cloud GPT/Claude, and specialized models. Best-fit per task, cost-optimized automatically. |
Single-provider bias Primarily tied to one cloud provider. Switching costs are high. |
| Security & Privacy | ||
| Data Residency |
User's device Personal context, documents, and behavior data never leave the PC. |
Cloud servers User prompts and documents sent to third-party APIs for processing. |
| Enterprise Compliance |
On-device = compliant by default No data egress. Simplifies GDPR, CCPA, and enterprise procurement. |
Requires cloud DPA Enterprise buyers must evaluate third-party data processing agreements. |
| Intelligence & Personalization | ||
| Context Engine |
Mnemosyne (Watch & Learn) Continuously learns from user behavior. Builds a personal knowledge graph on-device. Agents improve over weeks and months. |
Stateless by default Each session starts fresh. No built-in mechanism to learn from user patterns across sessions. |
| Accuracy Over Time |
Gets better with use More usage = better context = more accurate results. Hardware uptime directly improves AI quality. |
Static Quality depends on prompt engineering. No hardware-linked improvement loop. |
| OEM Business Model | ||
| Revenue to OEM |
~$545/user over 3 years (est.) Estimated breakdown: Data annotation (~$300) + Skills marketplace (~$80) + Subscription share (~$165). Recurring. Actual figures vary by OEM terms. |
$0 recurring Open-source framework. No built-in monetization path for OEM partners. |
| Hardware Story |
Hybrid compute = hardware matters Intelligent scheduling routes tasks to local NPU/GPU when possible, saving cloud costs. Better hardware = more tasks run locally = lower cost for users. |
No hardware tie-in Cloud-first approach means hardware specs are irrelevant to AI performance. |
| Ecosystem |
Open standards + curated marketplace Compatible with OpenClaw and agentskills.io standard. IrisGo Skills Marketplace lets creators monetize their Skills, with curated quality assurance for end users. |
Community-driven Growing open-source community. Quality and availability vary. No built-in commerce layer. |
| Time to Ship |
3-6 months Turnkey integration. IrisGo handles updates, support, and ecosystem management. |
12-18 months OEM must build: installer, update mechanism, user support, skill curation, billing. |
OpenClaw is a capable framework. IrisGo is what turns it into a business.
IrisGo's hybrid compute engine intelligently schedules tasks between on-device and cloud. More RAM, faster NPU = more tasks run locally = lower cost for users. That is a story your sales team can tell.
Enterprise and government buyers ask one question: "Where does my data go?" With IrisGo, the answer is simple: it stays on the device. No cloud DPA negotiations, no data egress concerns.
Watch & Learn (Mnemosyne) observes how users work and builds personal context on-device. After 30 days, IrisGo knows your work patterns, meeting habits, and document preferences. OpenClaw starts from zero every session.
IrisGo's three-engine model (Data Annotation + Skills Marketplace + Subscription) generates an estimated ~$545 per user over 3 years. OpenClaw generates $0 for OEMs.
OpenClaw is a framework. It gives developers building blocks to create AI agents. But OEMs still need to build the product, the ecosystem, the monetization, and the support layer.
IrisGo is a platform. Pre-built agents, on-device intelligence, a Skills Marketplace with 5,000+ Skills, and a revenue model that pays OEMs from day one.
They are complementary. IrisGo runs on the same open standards (agentskills.io compatible). We bring the managed experience, the business model, and the hybrid compute optimization that makes your hardware the hero.
Let's discuss how IrisGo's hybrid compute engine makes your hardware the AI differentiator.