Sovereign
Infrastructure
Cloud.
Zero hypervisor markup. Zero egress fees. Enterprise multi-tenant IAM with Argon2id and AWS SigV4, running directly on sovereign cloud infrastructure with high-performance compute and NVMe storage.
Connect Your AI Agent to Cloud Infrastructure
Plug `skiff-cloud` into Cursor, Claude Desktop, or your custom agent loop via the official Model Context Protocol (MCP) standard.
{
"mcpServers": {
"skiff": {
"command": "npx",
"args": ["-y", "skiff-cloud", "mcp"]
}
}
}Sovereign Cloud Primitives
Full cloud surface area guarded by Argon2id cryptographic identity and SigV4 verification.
Cockpit Telemetry
Live 60Hz SSE telemetry covering cloud CPU cores, DDR4 RAM, NVMe storage, and encrypted mesh routing.
Compute Sandboxes
Instant cloud OCI containers with automatic subdomain routing and cgroups v2 memory ceilings.
Data & SQL Studio
Managed PostgreSQL instances with pgvector HNSW indexing, alongside ultra-fast SiderDB LSM-tree caching.
GPU Studio
Infrastructure accelerated FP16 inference stream exposed via OpenAI-compatible endpoints.
AWS S3 Drop-In
Content-addressed NVMe object store supporting AWS SigV4 HMAC-SHA256 signatures and Boto3.
Multi-Tenant IAM
Argon2id password hashing, sovereign JWT tokens, hard infrastructure quotas, and instant SigV4 access keys.
# Authenticate with Sovereign Node $ npx skiff-cloud login Email: admin@skiff.local Password: •••••••••••• ✔ Successfully authenticated as usr_root_admin ✔ Session cached to ~/.skiff/config.json # Check current tenant & quotas $ npx skiff-cloud whoami Tenant: System Root (tenant_root) Role: admin Infrastructure: Dedicated Cloud Infrastructure # Deploy sandboxes or query SQL $ npx skiff-cloud ps $ npx skiff-cloud deploy --name api --image node:20-alpine $ npx skiff-cloud db query "SELECT * FROM documents LIMIT 5"
import boto3
# Connect to Sovereign S3 drop-in
s3 = boto3.client(
"s3",
endpoint_url="https://s3.skiffcloud.app",
aws_access_key_id="skf_akia_...",
aws_secret_access_key="skf_sec_...",
region_name="us-east-1",
)
# Upload object directly to NVMe
s3.upload_file("model.bin", "weights", "model.bin")