Self-Hosted LLM Proxy

Use Stockyard as just an LLM proxy.

Point your app at Stockyard and route requests to OpenAI, Anthropic, Google, Ollama, and 12 more providers through a single stable endpoint. No Redis, no Postgres, no Docker.

Install and proxy in 60 seconds

$ curl -fsSL stockyard.dev/install.sh | sh
✓ Stockyard installed (~25MB)

$ stockyard
Proxy: http://localhost:4200/v1
✓ 76 modules active
Console: http://localhost:4200/ui

# Use it like you'd use OpenAI directly
$ curl localhost:4200/v1/chat/completions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -d '{"model":"gpt-4o-mini","messages":[{"role":"user","content":"hello"}]}'
  

That request just passed through 76 middleware modules. Traced, logged, audited — automatically.

Change one line in your code

Python

client = OpenAI(
  base_url="http://localhost:4200/v1"
)
      

Node / TypeScript

const openai = new OpenAI({
  baseURL: "http://localhost:4200/v1"
});
      

Cursor / Windsurf / Copilot

# Editor settings → API Base URL:
http://localhost:4200/v1
      

LangChain / LlamaIndex / Any SDK

# Any OpenAI-compatible SDK works.
# Just change the base URL.
      

Full integration docs · Editor setup guides

What the proxy gives you immediately

Tracing on every request

Cost, latency, token count, model, and provider — recorded automatically. Query traces through the API or browse them in the dashboard.

Free — Lookout

Tamper-proof audit ledger

Every request gets a hash-chained ledger entry. You can prove what was sent, what came back, and that nothing was altered.

Free — Brand

Runtime controls

Rate limiting, token budgets, content filtering, PII redaction, prompt injection detection — all configurable per-route, all running in the middleware chain.

Free — 76 modules

Failover routing

If a provider goes down, Stockyard fails over to the next one automatically. Model-aware routing sends Claude requests to Anthropic first, GPT requests to OpenAI first.

Free — Chute

Model aliasing

Map stable names like "fast" or "smart" to real models underneath. Swap providers later without touching app code.

Free — Alias API

You don't have to use everything

Stockyard ships with 150 tools and 76 modules, but you don't need to configure any of them. Install it, point your app at it, and the proxy works out of the box. Tracing, auditing, and controls are automatic. Everything else is optional — turn it on when a real need shows up.

Most people start with just the proxy. That's the whole idea.

What you gain vs what you give up

What you gain: one stable endpoint for 16 providers, automatic request tracing and cost tracking, response caching, rate limiting, model aliasing, failover routing, and a web console. All from one binary with no external dependencies.

What you give up: roughly 200ms of added latency per request (the middleware chain). Your requests route through one more hop. If you need sub-millisecond proxy overhead or horizontal scaling across many nodes, a lighter proxy or direct provider calls may be a better fit.

No phone-home
Stockyard does not send telemetry, analytics, or usage data. Your requests go to the LLM providers you configure and nowhere else. Full details →

Example: proxy-only config

# stockyard.yaml -- proxy-only setup
# Just set your provider keys. Everything else is automatic.

providers:
  openai:
    api_key: ${OPENAI_API_KEY}
  anthropic:
    api_key: ${ANTHROPIC_API_KEY}

# Tracing, cost tracking, and audit logging are on by default.
# No modules to configure. No YAML to write beyond this.
  

OpenAI, Anthropic, Google Gemini, Mistral, Cohere, Groq, Together AI, Fireworks, Perplexity, DeepSeek, Ollama, LM Studio, Azure OpenAI, AWS Bedrock, OpenRouter, and any OpenAI-compatible endpoint.

Start with just the proxy.

One binary. One SQLite file. Every request traced. Install in under a minute, no credit card needed.

Install Stockyard

Proxy Docs → Model Aliasing → vs LiteLLM →

GitHub·Changelog·BSL 1.1

Frequently Asked Questions

Can I use Stockyard as just a proxy without the other features?

Yes. Proxy-only is a first-class use case. Install the binary, set your provider key, and route requests through one endpoint. Tracing, audit, and the other products are there if you want them later, but they are not required.

Does proxy-only mode cost anything?

No. The Community tier is free and includes the full proxy with all 76 middleware modules, 16 provider integrations, caching, rate limiting, and cost tracking. No credit card required.

What providers does the proxy support?

Stockyard supports 16 LLM providers including OpenAI, Anthropic, Google Gemini, Groq, Mistral, DeepSeek, Together AI, Fireworks, Ollama, and more. Set an environment variable and the provider is auto-configured.

Do I need Docker or Redis to run the proxy?

No. Stockyard ships as a single Go binary with embedded SQLite. No Docker, no Redis, no Postgres, no external dependencies. Download and run.