DeepSeek · OpenAI · Gemini · Claude — no local LLM needed

One Ollama endpoint.
Your favorite cloud AI.

AI Cloud Proxy is a free Windows app that exposes an Ollama-compatible server on your machine and forwards every request to the cloud provider you choose. Point your VS Code extension, Continue, Cline or any Ollama client at 127.0.0.1 — no model downloads, no GPU required.

Download for Windows Quick start →

✓ 100% local listener
✓ Free & open source
✓ Bring Your Own Model ready

127.0.0.1:11435 — Ollama API

$ curl http://127.0.0.1:11435/api/chat -d '{
  "model": "deepseek-v4-flash@DeepSeek",
  "messages": [{ "role": "user",
                 "content": "Hello!" }],
  "stream": true
}'

DeepSeek →

Hello! I'm a cloud model reached through an Ollama-compatible endpoint. 🎉

Routes to

DeepSeek
OpenAI
Gemini
Claude
+ any OpenAI-compatible

Features

Everything your Ollama workflow needs —
pointed at the cloud.

Ollama-compatible API

/api/chat, /api/generate and /api/show with NDJSON streaming — your existing Ollama clients keep working unchanged.

OpenAI-compatible too

/v1/chat/completions (SSE) and /v1/models for OpenAI-style tools and libraries.

Multi-provider routing

Every provider with a saved key is active at once. Models are advertised as model@Provider (e.g. gemini-2.5-flash@Gemini) and requests route to the provider that owns them.

Model auto-discovery

“Get Models” pulls the live model list from your provider and fills the dropdown. On start, your configured model is validated and you’re warned if it vanished.

Per-provider settings

Each provider keeps its own API key, base URL and last model. Switch providers and everything is restored automatically.

Built-in test tab

Send a question and watch the answer stream straight from the selected provider — perfect for verifying your setup in seconds.

Runs in the system tray

Close the window and it keeps running next to the clock. Reopen or exit from the tray icon anytime.

Custom providers

Add your own OpenAI-compatible endpoint (name + base URL + key + model) and it appears in the dropdown like any other provider.

Tip: put the full API path in the Base URL — e.g. https://openrouter.ai/api/v1 — because the app appends /chat/completions and /models to it.

Local & private

Keys and settings live in %APPDATA%\AiCloudProxy. No accounts, no tracking — requests go only to the provider you choose.

How it works

Translate once. Use everywhere.

AI Cloud Proxy speaks the Ollama wire format to your tools and each provider’s native format to the cloud — translating on the fly.

1 · Your tool

VS Code extension, Continue, Cline, curl… anything that speaks Ollama.

POST /api/chat

2 · AI Cloud Proxy

In-process HTTP listener that translates to each provider’s native format.

127.0.0.1:11435

3 · Cloud provider

DeepSeek, OpenAI, Gemini or Claude answers and streams back the same way.

provider API

Provider formats translated by AI Cloud Proxy (default models are editable)
Provider	Default model	Wire format
DeepSeek	`deepseek-v4-flash`	OpenAI-compatible
OpenAI	`gpt-4o-mini`	OpenAI-compatible
Gemini	`gemini-2.5-flash`	`generateContent`
Claude	`claude-sonnet-4-5-20250929`	Anthropic Messages
Custom	(yours)	OpenAI-compatible

💡 DeepSeek also exposes deepseek-v4-pro and the experimental deepseek-v4-flash-vision-exp (image input) — use Get Models in the app to fill the dropdown automatically.

Endpoints

The listener you can point anything at.

POST/api/chat

Ollama chat format — NDJSON streaming

POST/api/generate

Ollama generate format — NDJSON streaming

POST/GET/api/show

Per-model metadata + capabilities (VS 2026 Copilot BYOM)

POST/v1/chat/completions

OpenAI format — SSE streaming

GET/api/tags

All discovered models with capabilities & token limits

GET/v1/models

OpenAI-compatible model list

Quick start

Up and streaming in ~1 minute

1
Run AICloudProxy.exe — the app opens on the Proxy Server tab.
2
Pick a provider and click Get API Key ↗ to grab a key from that provider’s site.
3
Paste the key into the API Key box and set the Port (default 11435).
4
Click Start Proxy.
5
Point your Ollama tool at http://127.0.0.1:11435 (or your port) — done!
6
Use the Test tab to send a sample question and verify the stream.

Visual Studio 2026 Copilot (BYOM)

AI Cloud Proxy advertises tool-calling capabilities and token limits via /api/show and /api/tags, so VS 2026 Copilot lists its models. Enable Bring Your Own Key in Tools → Options, add AI Cloud Proxy as an Ollama provider at http://127.0.0.1:11435, then pick a model in Agent mode.

Screenshots

See it in action

AI Cloud Proxy main window: provider selection, API key, port, Start/Stop and a green Running badge — Main window — pick a provider (Gemini shown), paste your API key, and Start the proxy.

AI Cloud Proxy Test tab streaming a live answer from the selected provider — Test tab — ask a question and stream the answer live from your provider.

AI Cloud Proxy options, feedback panel and live log showing the active providers — Options, feedback & live log — active providers at a glance.

AI Cloud Proxy first-run Quick Tour dialog walking through the setup steps — Quick Tour — a 30-second guided setup on first launch.

AI Cloud Proxy system tray context menu — Open, Start/Stop Proxy, Exit — System tray — right-click the tray icon to reopen, start/stop the proxy, or exit.

FAQ

Questions, answered

Do I still need to install Ollama?

No. AI Cloud Proxy is a self-contained listener. If a tool expects an Ollama server, point it at AI Cloud Proxy instead — nothing else needs to be installed.

Which providers are supported?

DeepSeek, OpenAI, Gemini and Claude out of the box, plus any custom OpenAI-compatible endpoint you add. Multiple providers can be active at the same time.

Which tools work with it?

Anything that speaks the Ollama HTTP API: VS Code extensions, Continue, Cline, LangChain integrations, scripts and more. OpenAI-format tools work too.

Is my data private?

AI Cloud Proxy stores your keys and settings locally (no accounts, no telemetry). Chat requests go directly to the provider you selected — review that provider’s privacy policy for how they handle your prompts.

What port does it use?

The default is 11435, but you can change it to anything you like from the main window.

Stop downloading models. Start streaming.

Bring your favorite cloud AI into the tools you already use.

Download AI Cloud Proxy