Privacy-First · Source Available · BSL 1.1

Your LLM traffic.
Your rules.

The privacy-first LLM proxy and AI gateway that never sees your prompts.

Self-hosted gateway for OpenAI, Anthropic, Azure, Ollama, and vLLM - with load balancing, API key management, usage tracking, and rate limiting. Your infrastructure, zero knowledge of your data.

Star us on GitHub
void-terminal
$ curl http://localhost:8080/v1/chat/completions \
-H "Authorization: Bearer vl_uk_a3f2..." \
-d '{"model":"default","messages":[{"role":"user","content":"hello"}]}'
->

[Proxy] Routing to: anthropic (claude-sonnet)

[Auth] Key: "vl_uk_a3f2..." (Active)

[Fiber] Request handled in 1.2ms

Zero-Knowledge LLM Proxy

VoidLLM never stores, logs, or persists any prompt or response content. Not as a setting you can toggle - by architecture. This AI gateway is a zero-knowledge pass-through, GDPR compliant by design. Only metadata is tracked: who made the request, which model, how many tokens. Your data stays yours.

LLM Gateway Features

Everything you need to proxy, route, and control LLM API access across your organization.

Sub-2ms Proxy Overhead

In-memory auth and model resolution. Async usage logging. The LLM proxy adds under 2ms - the bottleneck is always the upstream provider, never VoidLLM.

Multi-Provider Routing

Use VoidLLM as an OpenAI proxy, Anthropic proxy, Azure OpenAI proxy, vLLM proxy, or Ollama proxy - all through one gateway. Load balance across deployments with automatic failover.

Full Web UI

Manage your LLM gateway from a built-in dashboard: playground, API key management, teams, models, usage tracking, and settings. No separate admin tool needed.

RBAC Built In

Org → Team → User → Key hierarchy with four roles. Most-restrictive-wins limits. Model access control per org, team, and key.

Token Budgets & Rate Limiting

Daily and monthly token limits. Requests per minute and per day. Set rate limits at org, team, or key level - enforced in real-time by the proxy.

Self-Hosted, Single Binary

Deploy the entire LLM proxy as one Go binary (~25MB). SQLite default, PostgreSQL optional. Docker image, Helm chart, or just go run.

Your AI gateway,
ready in minutes.

01

Deploy

Self-host VoidLLM with Docker, Helm, or go build. One binary, zero external dependencies. Runs on any cloud or on-prem hardware.

02

Configure

Add LLM providers via YAML or the Admin API. Create teams, manage API keys, set rate limits and token budgets.

03

Proxy

Point your apps at the proxy. Any OpenAI-compatible SDK works - just change the base URL and start routing.

LLM Proxy Pricing

Choose the plan that fits your team. Every tier includes the full proxy and AI gateway.

Flat pricing - no per-user fees, no per-request charges.

Save 2 months with annual billing: $490/yr Pro, $1,490/yr Enterprise

Community

Free

Free forever

  • 1 organization, 3 teams
  • Unlimited users
  • Multi-provider proxy (6 providers)
  • Full Web UI + Playground
  • Rate limiting + token budgets
  • Circuit breakers + per-model timeouts
  • Load balancing + automatic failover
  • MCP server (Claude Code, Cursor, Windsurf)
  • Prometheus metrics
  • Usage export (CSV/JSON)
  • BSL 1.1 - free to self-host
Get Started
Most Popular

Pro

$49 /mo

For teams in production

  • Everything in Community, plus:
  • Unlimited organizations + teams
  • Cost reports + budget alerts
  • Cross-org analytics
  • Priority email support (48h)
Start Free Trial

Enterprise

$149 /mo

Organization-wide control

  • Everything in Pro, plus:
  • SSO / OIDC (any provider)
  • Per-org SSO configuration
  • Audit logs (API + UI)
  • OpenTelemetry tracing
  • Multi-instance (Redis)
  • Dedicated Slack support (24h)
Start Free Trial
Limited spots

Founding Member

$999 one-time

Enterprise forever

  • All Enterprise features - current and future
  • Lifetime license - no recurring fees
  • Product Advisory Board membership
  • Direct founder access (Slack)
  • Priority support
  • Early access to new features
Become a Founding Member

Common questions

See all →

Does VoidLLM log my prompts?

No. There is no content logging code. Only metadata is tracked. Read about our zero-knowledge architecture.

Is VoidLLM free?

The Community tier is free with no time limit. Load balancing, MCP, RBAC, and the full admin UI included.

Does VoidLLM support load balancing?

Yes. Four strategies with automatic failover and circuit breakers. More on load balancing and failover.

Can I use VoidLLM with Claude Code or Cursor?

Cursor and Windsurf as LLM proxy. Claude Code via MCP. Connect Cursor, Windsurf, and Claude Code.

From the blog

View all →