Resource pool
GemmaNode works with a broad, extensible pool of resources. Availability, quotas, and pricing for any third-party resource are set by that provider and can change — nothing here is a permanent guarantee.
Local Compute
Run workloads on your own hardware — private, low-latency execution when available.
- Your CPU / GPU / RAM
Cloud Compute
Draw on compatible remote compute when extra capacity is useful, subject to each platform's own limits.
- Camber
- Kaggle
- Google Colab
- Hugging Face
AI Providers
Use external AI APIs for suitable workloads, respecting each provider's own terms and quotas.
- Gemini API
- OpenRouter
- Pollinations AI
Autonomous Agents
Multi-step, tool-using execution for tasks that need more than a single model call.
- OpenHands
- OpenClaw
- Hermes
Manual Escalation
Complex or uncertain tasks can optionally be escalated to a stronger model for human-reviewed handling.
- Claude
Built on free & user-controlled resources
GemmaNode is designed to make the most of free tiers, open-weight models, and your own hardware — not to promise unlimited free compute forever.
Free-tier aware
Designed to make good use of free and low-cost resources, while respecting each provider's own limits and policies.
BYOK
Bring your own keys where a provider requires one — nothing shared, nothing hardcoded.
No forced subscription
The core, self-hosted experience isn't metered by GemmaNode itself.
Your hardware counts
Local compute is a first-class resource, not an afterthought — useful even without a dedicated GPU.
Provider limits still apply
Free tiers, quotas, and pricing are set by each provider and can change at any time.
No hidden metering
You can see which resource handled a task — no silent switch to a paid tier without your setup allowing it.
Built to evolve with the AI ecosystem
GemmaNode's provider and agent boundaries are designed to be swappable, not hardcoded.
New models
Additional models can be added to the pool as they become useful.
New AI providers
Providers beyond the current set can be integrated over time.
New compute resources
Additional local or cloud compute sources can be supported.
New agent frameworks
The agent layer isn't limited to what's currently listed.
Open source
Fully inspectable code, free to fork, audit, and extend.
No black boxes
What's public stays high-level by design — but nothing is hidden about how to contribute.