One gateway for every model, agent, and workflow.
ProxyLLM is an OpenAI-compatible endpoint. Anything that lets you set a base URL works with it: SDKs, automation platforms, coding agents, LLM routers, and model providers. OpenAI-bound work runs through Codex Hosted on your ChatGPT subscription.
n8n
Use the built-in OpenAI credential. Docker workflows get Codex Hosted and budget caps.
Make
Automation scenarios get one OpenAI-compatible credential and a hard spend guardrail.
Zapier
Webhooks by Zapier posts to one OpenAI-compatible endpoint, behind a scoped key and a monthly cap.
Dify
App-builder prompts go through ProxyLLM with request logs and fallback policy.
Flowise
Visual agent flows get scoped keys and budget caps without custom nodes.
GitHub Actions
CI evals and review jobs run with capped keys and request-level usage logs.
Node.js
Use the official OpenAI SDK with a ProxyLLM base URL and per-service keys.
Python
Swap `base_url` once for scripts, jobs, notebooks, and eval runners.
Vercel AI SDK
Use the OpenAI-compatible provider and keep your app code unchanged.
LangChain
Put chains and agents behind one gateway with run-level spend visibility.
LlamaIndex
Point query engines and RAG jobs at one OpenAI-compatible client.
REST API
Plain `/v1/chat/completions` for curl, workers, edge functions, and custom clients.
OpenAI
Keep your OpenAI models and SDK. Run the work on your ChatGPT subscription, keep your key as fallback.
Claude
Sonnet, Haiku, and Opus behind the same OpenAI-compatible endpoint and budget controls.
Gemini
Flash for cheap high-volume work, Pro for harder tasks, with one ProxyLLM key.
DeepSeek
Cheap reasoning and code models for agent loops, with spend caps per app or workflow.
xAI / Grok
Grok models for live-style assistant workloads without exposing provider keys everywhere.
Meta Llama
Open-weight Llama models for cheap summarization and classification behind one endpoint.
Mistral
Fast European models with app-level usage controls and request logs in front.
Cohere
Enterprise text and retrieval-heavy workloads tracked through one gateway.
Groq
Very fast inference for interactive agents and latency-sensitive work.
Perplexity
Search-answer models with separate keys, budgets, and request logs.
Together
Open model catalog access with ProxyLLM keys and cost visibility.
Fireworks
Production open-model serving behind scoped app keys.
Cerebras
High-throughput inference for fast agent steps and batch jobs.
Replicate
Hosted open models and specialized workloads behind one auditable endpoint.
Hugging Face
Inference endpoints and open models tracked alongside the rest of your LLM traffic.
Azure OpenAI
Enterprise OpenAI deployments with centralized keys, budgets, and logs.
Amazon Bedrock
AWS-hosted model access framed as a controlled enterprise provider lane.
Vertex AI
Google Cloud model deployments with ProxyLLM keys and governance in front.
opencode
One provider block in opencode.json. Sessions run on your ChatGPT subscription.
Cursor
Put coding-agent traffic behind scoped keys and budget caps.
Codex
Connect your ChatGPT account and run Codex hosted, around the clock, from any app.
MCP
Expose usage and key controls to agent clients as tools.
Run your AI workloads on your ChatGPT subscription.
ProxyLLM runs OpenAI's Codex for you, signed in with your own ChatGPT account. Your apps call one OpenAI-compatible endpoint and the work bills to your flat plan instead of per-token API pricing.
Missing your tool? It probably still works.
LangChain, Cursor, Vercel AI SDK, plain curl: if it accepts an OpenAI base URL, it accepts ProxyLLM. Set the base URL, use your ProxyLLM key, done.