Overview
VCodeX is the control plane for AI apps — one OpenAI-compatible gateway to route, govern and observe every model.
VCodeX puts a single control plane in front of every model your team uses. Point your existing OpenAI SDK — or an editor like Claude Code, Cursor or Zed — at the gateway, connect your providers, set guardrails and budgets once, and ship with full observability across all 17 providers.
Base URL
The gateway lives at https://app.vcodex.ai/v1 (OpenAI-compatible) and
https://app.vcodex.ai/v1/messages (Anthropic-native). Everything above
assumes a client pointed at one of those with a VCodeX sk-... key.
Where to start
- Connecting an AI tool or IDE? → For AI Tools & IDEs (Claude Code, opencode, Cursor, Zed, Continue, aider).
- Building an app against the gateway? → For Developers (OpenAI-compatible API, Anthropic Messages, streaming, models).
- Setting up the platform for your org? → Bring your own keys, Budgets & quotas, Guardrails.
What you get
- AI Gateway — one OpenAI-compatible endpoint (
/v1/chat/completions) for 17 providers, with streaming, fallbacks and an Anthropic Messages API. - Auto-mode routing —
vcodex/autoscores each prompt and routes to an economical, balanced or powerful model. - Bring your own keys — your provider keys, encrypted in AWS Secrets Manager, resolved per request.
- Budgets & quotas — per-user and per-org budgets enforced in real time, with alerts at 80 / 90 / 100%.
- Observability — every request logged with tokens, cost, latency and a ULID correlation ID.
- Guardrails & governance — PII redaction, secret blocking, RBAC, IP rules and audit logs.