What is MoltGhost
MoltGhost is a platform for deploying and running AI agents in fully isolated compute environments.
Unlike shared runtimes, MoltGhost provisions a dedicated Agent Pod for each agent, containing all necessary compute resources, runtime software, and language models for independent operation.
Agent-Based Infrastructure
MoltGhost centers the agent as the primary system unit.
When deploying an agent, the platform automatically provisions an Agent Pod with:
- Compute resources (GPU/CPU)
- Agent runtime framework
- Local language model
- Integrated tools
This enables agents to process inputs, perform reasoning, execute tools, and generate outputs autonomously.
Deploy Agent → Provision Agent Pod → Load Model & Runtime → Agent Active
Independent Agent Environments
Every agent operates in complete isolation:
Agent A → Agent Pod A (CPU/GPU + Runtime + Model A + Data A)
Agent B → Agent Pod B (CPU/GPU + Runtime + Model B + Data B)
Agent C → Agent Pod C (CPU/GPU + Runtime + Model C + Data C)
Key Benefits:
- No shared runtime processes
- Isolated model execution
- Separated agent data and state
- Independent scaling and failure domains
Local Model Execution
MoltGhost agents run language models locally within their Agent Pod using an optimized model runtime.
Advantages:
- No external API dependencies
- Full control over inference parameters
- Predictable latency and compute costs
- Custom model fine-tuning support
Agent Pod → Local Model Runtime → LLM Inference → Agent Response
Managed Agent Access
Deployed agents receive secure, platform-managed endpoints:
https://<agent-id>.agent.moltghost.io/v1/chat
https://<agent-id>.agent.moltghost.io/v1/tools
Users interact via standard HTTP APIs without direct compute access, ensuring security and simplicity.
Agent Lifecycle Management
MoltGhost provides full control over agent operations:
| Operation | Description | Use Case |
|---|---|---|
| Deploy | Create & start new Agent Pod | Launch production agent |
| Start | Activate paused Agent Pod | Resume after maintenance |
| Stop | Pause & preserve state | Reduce costs during idle periods |
| Resume | Restart from last state | Continue interrupted work |
| Terminate | Delete Agent Pod & resources | Permanent cleanup |
Example CLI Operations
moltghost agent deploy my-agent --model llama3.1-70b
moltghost agent stop my-agent
moltghost agent resume my-agent
Summary
MoltGhost transforms AI agents into independent, production-ready compute environments.
Core Value Proposition:
✅ Per-agent isolation via dedicated Agent Pods
✅ Local model execution without API dependencies
✅ Automated lifecycle management
✅ Secure, scalable agent endpoints
Build, deploy, and scale agents as easily as containerized applications—with complete resource isolation.