Under the Hood
Three layers. One platform.
Routing Engine
Stateless request router. Validates your API key, selects the right provider channel based on your routing rules, and forwards the request. Streams the response back. No content stored. OpenAI-compatible endpoint.
70+
model aliases
<5ms
routing overhead
4
providers
- Smart routing via
1k4-custom: configure rules per content type (code, reasoning, images) - Per-project API keys with independent routing rules
- Works with any OpenAI SDK client
Agent Runtime
The Lab workspace runs a full agent runtime: multi-step task execution, tool calling, file management, document editing, code review, parallel subagents, and compute orchestration.
45+
skills
Infinite
context
Parallel
agents
DOCX/PDF
editing
- Context management: agent handles its own window, no re-explaining
- Adaptive complexity: adjusts approach based on task difficulty
- Local models via Bridge: same agent capabilities on your hardware
- Compute: provision VMs, deploy code, configure domains
Account & Billing
Prepaid credits. $1 = 1 credit. No subscriptions, no overage charges. Per-token billing for text, per-unit for media. Local models are free.
- Real-time balance tracking with per-request deduction
- Per-project usage breakdown
- Automatic billing protection: requests stop at zero balance (no debt)
- Credits never expire
What you gain
One key
Single API key for every provider. One bill.
Intelligent routing
Content-aware model selection. Or choose your own.
Agent workspace
Not a chatbot. A workspace that executes.
Local models
Bridge your Ollama. Full agent power. $0.
Compute
Real servers, real deployments, real URLs.
Claude Code review
Multi-model plan review via MCP. Integrated.