Your local AI workstation, no cloud required.

Multi-Engine Runtime

llama.cpp (CPU), MLX (Apple Silicon), vLLM (CUDA), and ONNX — one unified adapter interface. Switch backends without changing application code.

OpenAI-Compatible API

Drop-in `/v1` endpoint. Any SDK or tool targeting OpenAI works unchanged. Anthropic-compat at `/anthropic/v1` for Claude-style tool use.

RAG + Knowledge Bases

Qdrant vector store, Meilisearch full-text, per-workspace isolation. Ingest PDFs, DOCX, Markdown, and plain text with async chunking pipelines.

Agent Orchestration

Compose agents with tool bindings, MCP server allowlists, and memory. Code-exec sandbox with vm timeout, memory cap, and egress allowlist.

Workflow Automation

Visual DAG editor. Cron, webhook, model-event, and chat-event triggers. BullMQ-backed execution with per-step progress streaming.

Enterprise Security

Postgres RLS tenancy, constant-time API-key compare, gateway loopback-only bind, per-key scopes, audit log (append-only, tamper-proof).

5

Inference engines

31

Database tables

100

Build batches

14

Packages

< 50 ms

API gateway p50

Local

Free

Single user, local machine only.

• All inference engines
• OpenAI-compat API
• Chat + Prompt Studio
• Local RAG
• Community support

Pro

$12/mo

Power users and solo developers.

• Everything in Local
• Remote model sync
• Agents + Workflows
• Priority support
• Marketplace access

Team

$49/mo

Up to 10 seats, shared workspace.

• Everything in Pro
• Multi-user workspace
• Role-based access
• SSO (OIDC)
• Audit logs

Enterprise

Custom

Air-gapped, unlimited seats.

• Everything in Team
• Air-gap mode
• Custom SLA
• On-prem deployment
• Dedicated support