OpenBlue

Your Local AI Gateway for VS Code

Run any open-source LLM as a first-class provider inside Visual Studio Code — through GitHub Copilot Chat, Claude Code, or the native Language Model API.

Download · Quick Start · Features · Compatibility · FAQ · License

---

Status Bar Tooltip Hover the polar bear icon — live status without opening anything.

Main Menu (Quick Pick) Click the icon — full management surface in a single Quick Pick. Ollama Model Downloader Search and pull models from the Ollama registry without leaving the editor.

---

🌊 What is OpenBlue?

OpenBlue is a Visual Studio Code extension that turns your local model runtimes — LM Studio and Ollama — into native providers inside the IDE. Your locally hosted models appear in VS Code's native model picker, in GitHub Copilot Chat, in Claude Code, and as inline completions. No cloud round-trips. No API keys. No data ever leaves your machine.

One bridge, every model, zero telemetry.

OpenBlue is an independent product. It does not include, install, or redistribute any third-party software — it simply talks to the local HTTP APIs of runtimes you already own.

🏠 100% Local	🔓 Model Freedom	💰 Zero Cloud Cost	📴 Works Offline
Your code never leaves your network.	Any GGUF, any Hugging Face model, any Ollama tag.	Run on your own hardware — pay nothing per token.	Air-gapped, sovereign, fully autonomous.

---

📚 Table of Contents

• Why OpenBlue?
• Features
• Compatibility Matrix
• System Requirements
• Installation
• Quick Start
• Using OpenBlue
• User Interface Tour
• Configuration Reference
• Architecture
• Privacy & Security
• Troubleshooting
• FAQ
• License
• Third-Party Notices

---

💡 Why OpenBlue?

The current AI-coding landscape forces an uncomfortable choice:

Option	Privacy	Cost	Flexibility	Quality
Cloud-only assistants (Copilot, Claude)	❌ Sends code to vendor	💸 Subscription / per-token	🔒 One model family	⭐⭐⭐⭐⭐
Plain local LLM in a separate window	✅ 100% local	✅ Free	⚠️ No IDE context	⭐⭐
OpenBlue + Local Models	✅ 100% local	✅ Free	✅ Any model and IDE integration	⭐⭐⭐⭐

OpenBlue gives you the IDE integration of a cloud assistant with the privacy and freedom of running your own models. Drop in a Qwen Coder, a DeepSeek, a GLM, or a freshly-fine-tuned LoRA — and it just works in Copilot Chat and Claude Code as if it were native.

---

✨ Features

🔍 Auto-Discovery & Native Model Picker

OpenBlue automatically discovers every model loaded in LM Studio or pulled into Ollama and registers them in VS Code's native model picker. No manual config. Load a model in LM Studio → it appears in the picker. Pull a model with ollama pull → it's there too.

🤖 Three Integration Surfaces

Surface	What It Does	Activation
GitHub Copilot Chat	Local models appear as selectable providers in the Copilot Chat model dropdown. Full agent mode, ask mode, plan mode.	Automatic on first launch
Claude Code	Routes Claude's Haiku / Sonnet / Opus roles to your local models. Use the Claude Code CLI or panel — answers come from your GPU.	Optional, one-toggle
Inline Completions	Tab-autocomplete powered by FIM-capable models (qwen2.5-coder, codellama:code, deepseek-coder).	Optional, opt-in
Chat Participant	A @openblue participant in VS Code Chat for direct conversation with the underlying model.	Always available

📦 Full Model Management Inside VS Code

A single status-bar menu gives you a symmetric set of operations across both backends:

• 🔎 Search the Hugging Face GGUF catalog (LM Studio) or the Ollama registry (Ollama)
• ⬇️ Pull / Download new models without leaving VS Code
• 🚀 Load a model into VRAM (and let OpenBlue spawn it if needed)
• 💤 Unload to free memory
• 📋 List loaded vs. available
• 🗑️ Delete to reclaim disk space

🎛️ Per-Role Model Mapping

OpenBlue maps Claude Code's three logical roles — Haiku (fast), Sonnet (balanced), Opus (heaviest) — to specific local models of your choosing. Use a 4B for cheap edits, a 14B for normal tasks, and a 70B MoE for the hard ones, all selected automatically by the Claude Code agent's planner.

👁️ Visibility Control

Hide models you don't want to see in the picker (older quants, abandoned experiments). One click to reset.

🌐 Optional Web Search

Toggle web-search tools (web_search / web_fetch for Ollama, ephemeral MCP for LM Studio) on or off per session — keep your sandbox airtight, or let the agent reach the live web when needed.

🔌 Local MCP Server Support

Configure Model Context Protocol servers and OpenBlue forwards tool calls correctly through whichever backend you're using. Your local models gain filesystem, terminal, browser, and any other MCP-exposed capability.

⚡ Inline Completion (FIM)

A dedicated low-latency completion provider built on Fill-In-the-Middle prompting:

• Configurable debounce (default 300ms)
• Tunable temperature, max tokens, context window
• Custom stop sequences
• Auto-picks Ollama or LM Studio (or you choose explicitly)

🐻‍❄️ Status Bar Polar Bear

A single status-bar icon shows a tooltip with:

• Connection status for each backend.
• Available models.
• Claude routing status.
• Quick links to main actions.

Click to open the full management menu.

---

🧩 Compatibility Matrix

Supported IDEs / Editors

IDE / Editor	Status	Notes
Visual Studio Code	✅ Primary target	Stable
VSCodium	✅ Compatible	Tested on VSCodium 1.104+
Cursor	✅ Compatible	Inherits VS Code extension API — most features work
Antigravity	✅ Compatible	Inherits VS Code extension API — most features work
Every other IDE	Partially compatible	Inherits global claude config so you can use claude CLI with local open blue models

Supported Local Model Runtimes

Runtime	Versions	Auto-Detect	Auto-Spawn	Notes
LM Studio	0.3.0+	✅	✅ (configurable)	Native REST + WS API. GGUF models from HF.
Ollama	0.1.30+	✅	✅	Pull from registry, full model library access.
Ollama Cloud	Latest	✅	—	Sign-in detected automatically; routes to cloud-hosted models when chosen.
LM Studio Hub	Latest	✅	—	Sign-in detected; access to Hub-only catalog.

Supported Chat Surfaces

Surface	Local Models	Inline Completions	Tools / MCP
VS Code native model picker	✅	—	✅
GitHub Copilot Chat	✅	—	✅
Claude Code Chat	✅ via routing	—	✅
Claude Code CLI	✅ via env vars	—	✅
Editor inline (FIM)	—	✅	—

---

💻 System Requirements

Minimum

• OS: Windows 10/11, macOS 12+, or modern Linux (glibc 2.31+).
• VRAM: 8 GB for < 9B models + context, but >= 12 GB recommended for context viability.
• Disk: Depends on model size — at least > 10 GB free to start.
• VS Code: 1.104.0 or newer.
• Backend: LM Studio 0.3+ and/or Ollama 0.1.30+ installed and reachable.

Recommended

• OS: Windows 11, macOS 14+, Ubuntu 22.04+
• GPU: NVIDIA RTX 4070 / Apple Silicon M3+ / AMD RX 7800 XT.
• VRAM: > 32 GB comfortably runs up to 32B dense or MoE + 1M tokens context.
• RAM: 64 GB so swapping doesn't kill the pc when loading 80B coding models.
• Disk: SSD with 1 TB free for a healthy model library.

For The Best Coding Experience

NVIDIA DGX SPARK: Best price-performance for high level local inference.

---

📦 Installation

1. Download the latest release

   • Grab openblue-standalone-X.Y.Z.vsix from the dist/ folder of this repository
   • Or get it from the Releases page

2. Install in VS Code

   • Open the Extensions view (Ctrl+Shift+X)
   • Click the ··· menu → "Install from VSIX..."
   • Select the downloaded .vsix

3. Reload the window (Ctrl+Shift+P → "Developer: Reload Window") and follow the guided IDE installation steps.

---

🚀 Quick Start

1️⃣ Set Up a Backend

You can use LM Studio, Ollama, or both at once.

👾 LM Studio

1. Download from lmstudio.ai and install.
2. Open LM Studio → Local Server tab → Start Server.
3. In server settings, enable all options except Require Auth Token (unless you want one).
4. (Optional) Download a model now — or do it from inside VS Code later.

Default endpoint: http://localhost:1234

🦙 Ollama

1. Download from ollama.com and install.

2. Ollama starts automatically as a system service on port 11434.

3. (Optional) Pull a model from the terminal:

   ollama pull qwen2.5-coder:7b

Default endpoint: http://localhost:11434

2️⃣ Configure OpenBlue

After installing the extension:

1. Reload VS Code (Ctrl+Shift+P → "Developer: Reload Window")
2. Look for the 🐻‍❄️ polar bear icon in the status bar (bottom-right)
3. Click it → the OpenBlue main menu opens
4. (If prompted) Configure your backend endpoint — defaults usually just work:
   • LM Studio: http://localhost:1234
   • Ollama: http://localhost:11434
   • Or point to a remote machine: http://192.168.1.42:1234
5. Click Test Connection → blue means you're connected, green means you're Ollama or LM Studio logged.

3️⃣ Pick a Model & Start

With GitHub Copilot Chat:

1. Open the Chat panel (Ctrl+Shift+I)
2. Click the model dropdown at the bottom of the chat input
3. Select any of your local models (they appear under "OpenBlue (Local)")
4. Ask away — the request goes straight to your hardware

With Claude Code:

1. Open the OpenBlue menu → "Configure Claude Code Routing"
2. Map each role (Haiku / Sonnet / Opus) to a local model
3. Open Claude Code (panel or CLI) — it now uses your local models
4. Toggle routing on/off anytime from the status bar

With Inline Completions:

1. OpenBlue menu → "Toggle Inline Completions"
2. Pick a FIM-capable model (e.g. qwen2.5-coder:1.5b — fast and small)
3. Start typing in any file — completions appear after you pause

---

🎮 Using OpenBlue

The Status Bar Menu

Hover the 🐻‍❄️ icon for a live tooltip with backend status, available models, and quick actions. Click it to open the full contextual menu.

Status Bar Tooltip Hover the polar bear icon — live status without opening anything.

Main Menu (Quick Pick) Click the icon — full management surface in a single Quick Pick. Ollama Model Downloader Search and pull models from the Ollama registry without leaving the editor.

Available Commands

All commands are available from the Command Palette (Ctrl+Shift+P):

Command	Description
OpenBlue: Show Menu	Open the main contextual menu
OpenBlue: Open Chat (Ctrl+Shift+I)	Open the chat panel with @openblue selected
OpenBlue: Configure Server Connection	Edit the LM Studio / Ollama endpoint
OpenBlue: Pull Ollama Model	Download a new model from the Ollama registry
OpenBlue: Delete Ollama Model	Remove a local Ollama model
OpenBlue: Show Running Ollama Models	List models currently held in memory
OpenBlue: Browse Ollama Library	Search the Ollama registry inline
OpenBlue: Toggle Web Search	Enable / disable web-search tools
OpenBlue: Toggle Inline Completions	Enable / disable Tab autocomplete
OpenBlue: Toggle Claude Code Routing	Enable / disable Claude routing
OpenBlue: Configure MCP Servers	Open the MCP servers JSON config
OpenBlue: Repair Claude Code ServiceWorker	Fix Claude Code's local SW if it gets stuck
OpenBlue: Check Status	Print full diagnostic info to output channel

Keyboard Shortcuts

Shortcut	Action
Ctrl+Shift+I / Cmd+Shift+I	Open OpenBlue chat

You can rebind any command via VS Code's keyboard shortcuts editor.

---

🎨 User Interface Tour

1. The Status Bar Bear

A discreet 🐻‍❄️ icon sits in the bottom-right of VS Code. Color and dot indicators show real-time backend health.

2. The Main Menu (Quick Pick)

A native VS Code Quick Pick — no custom WebView clutter. Submenus drill down for granular control. Every entry has an iconographic glyph and a one-line description.

3. Inline Model Browser

When you choose "Browse Ollama library" or "Browse Hugging Face GGUFs", OpenBlue opens a searchable Quick Pick populated live from the corresponding registry. Pull, load, and start using a model in under a minute.

4. Connection Webview

The connection page (opened by clicking the connection entry) is a dark-themed config card — endpoint, port, status, and a one-click Test Connection button. It auto-fills with sensible defaults.

5. Model Mapping Editor

For Claude Code routing, a dedicated three-row picker lets you assign a local model to each Claude role. Changes are applied immediately.

6. Output Channel

Ctrl+Shift+U → select "OpenBlue" in the dropdown for live logs of every request, model load, and configuration write. Great for debugging.

---

⚙️ Configuration Reference

All settings live under the openblue.* namespace in settings.json.

Backend Endpoints

Setting	Type	Default	Description
openblue.lmStudio.host	string	http://localhost:1234	LM Studio server URL
openblue.lmStudio.executablePath	string	""	Path to LM Studio binary (empty = auto-detect)
openblue.lmStudio.maxInstancesPerModel	number	1	Max concurrent LM Studio instances per model
openblue.lmStudio.signedIn	boolean|null	null	Manual override for Hub sign-in detection
openblue.ollama.host	string	http://localhost:11434	Ollama server URL
openblue.ollama.executablePath	string	""	Path to Ollama binary (empty = auto-detect)
openblue.ollama.enabled	boolean	true	Enable Ollama integration
openblue.ollama.signedIn	boolean|null	null	Manual override for Ollama Cloud sign-in

Ollama Sampling Parameters

Setting	Type	Default	Description
openblue.ollama.topK	number	0	top_k (0 = disabled)
openblue.ollama.topP	number	0	top_p / nucleus (0 = disabled)
openblue.ollama.repeatPenalty	number	0	repeat_penalty (0 = disabled)
openblue.ollama.keepAlive	string	""	keep_alive duration (e.g. 5m, 1h, -1)

Claude Code Routing

Setting	Type	Default	Description
openblue.claudeRouting.enabled	boolean	true	Route Claude Code through OpenBlue
openblue.modelMapping.haiku	string	""	Local model ID for Haiku role
openblue.modelMapping.sonnet	string	""	Local model ID for Sonnet role
openblue.modelMapping.opus	string	""	Local model ID for Opus role
openblue.claudeCodeVersion	string	2.1.79	Pinned Claude Code CLI version

Inline Completions

Setting	Type	Default	Description
openblue.inlineCompletion.enabled	boolean	false	Enable Tab autocomplete
openblue.inlineCompletion.backend	enum	auto	auto / ollama / lmstudio
openblue.inlineCompletion.model	string	""	Override model ID (else backend-specific)
openblue.inlineCompletion.ollamaModel	string	qwen2.5-coder:1.5b	Default Ollama FIM model
openblue.inlineCompletion.lmStudioModel	string	""	Default LM Studio FIM model
openblue.inlineCompletion.debounceMs	number	300	Debounce after keystroke (0–2000)
openblue.inlineCompletion.maxTokens	number	256	Max tokens per completion (16–2048)
openblue.inlineCompletion.temperature	number	0.2	Sampling temperature (0–2)
openblue.inlineCompletion.maxContextLines	number	200	Context window around cursor (10–1000)
openblue.inlineCompletion.stop	string[]	["\n\n", "<|endoftext|>", …]	Stop sequences

Misc

Setting	Type	Default	Description
openblue.autoConfigureOnStartup	boolean	true	Auto-write compatibility env vars on launch
openblue.webSearch.enabled	boolean	false	Enable web_search / web_fetch tools

---

🏗️ Architecture

High-Level Diagram

%%{init: {'theme':'dark','themeVariables':{'fontFamily':'ui-monospace, SFMono-Regular, monospace','primaryColor':'#0d1117','primaryBorderColor':'#0077B6','primaryTextColor':'#e6edf3','lineColor':'#58a6ff','clusterBkg':'#0d1117','clusterBorder':'#30363d'},'flowchart':{'curve':'basis','nodeSpacing':45,'rankSpacing':60,'padding':18}}}%%
flowchart TB

    subgraph Surfaces["💬 Chat Surfaces"]
        direction LR
        CopilotChat["GitHub<br/>Copilot Chat"]:::surface
        ChatAt["@openblue<br/>Participant"]:::surface
        ClaudePanel["Claude Code<br/>Panel &amp; CLI"]:::surface
    end

    subgraph OB["🐻‍❄️ OpenBlue Extension"]
        direction TB
        subgraph Providers["VS Code Providers"]
            direction LR
            CMP["Chat Model<br/>Provider"]:::core
            ICP["Inline<br/>Completions"]:::core
            SBM["Status Bar<br/>Manager"]:::core
        end
        subgraph Bridges["Backend Bridges"]
            direction LR
            BM["Backend Managers<br/>LM Studio · Ollama"]:::bridge
            CCM["Claude Code Manager<br/>MCP · Routing"]:::bridge
        end
    end

    subgraph Runtimes["⚙️ Local Runtimes"]
        direction LR
        LMS["LM Studio<br/>:1234"]:::external
        OLL["Ollama<br/>:11434"]:::external
        CCC["Claude Code<br/>CLI binary"]:::external
    end

    GPU(["🎮 Your GPU / CPU memory"]):::hw

    CopilotChat --> CMP
    ChatAt --> CMP
    ClaudePanel --> CCM

    CMP --> BM
    ICP --> BM
    SBM --> BM
    SBM --> CCM
    BM --> CCM

    BM --> LMS
    BM --> OLL
    CCM --> CCC

    LMS --> GPU
    OLL --> GPU

    click CMP "https://github.com/SIRGPrice/OpenBlue#-features" _self
    click ICP "https://github.com/SIRGPrice/OpenBlue#-features" _self
    click SBM "https://github.com/SIRGPrice/OpenBlue#-user-interface-tour" _self
    click BM "https://github.com/SIRGPrice/OpenBlue#-compatibility-matrix" _self
    click CCM "https://github.com/SIRGPrice/OpenBlue#-features" _self
    click LMS "https://lmstudio.ai" _blank
    click OLL "https://ollama.com" _blank
    click CCC "https://anthropic.com" _blank

    classDef core fill:#0077B6,stroke:#90E0EF,stroke-width:2px,color:#ffffff
    classDef bridge fill:#023E8A,stroke:#00B4D8,stroke-width:2px,color:#ffffff
    classDef surface fill:#1f6feb,stroke:#a5d6ff,stroke-width:2px,color:#ffffff
    classDef external fill:#161b22,stroke:#58a6ff,stroke-width:2px,color:#e6edf3
    classDef hw fill:#0d1117,stroke:#3fb950,stroke-width:2px,color:#7ee787

Loading

How a Request Flows

1. User asks a question in Copilot Chat or Claude Code.
2. Surface (Copilot / Claude / native LM API) invokes the OpenBlue chat-model provider.
3. OpenBlue picks the backend (LM Studio or Ollama, depending on the chosen model).
4. The request is streamed to the backend's local HTTP endpoint.
5. Tokens are streamed back to the original surface — appearing in the chat UI in real time.
6. If tool calls are involved, OpenBlue translates them between the chat-surface's tool spec and the backend's tool spec (Ollama tools / LM Studio MCP).

Where Things Get Written

OpenBlue writes configuration in well-defined, documented locations only:

Path	Purpose	Reversible?
~/.claude.json (or platform equivalent)	Claude Code env vars for routing	✅
~/.openblue/config.json	Per-user OpenBlue state	✅
VS Code settings.json	User-visible settings	✅

Turning off Claude routing removes all OpenBlue-managed entries. No leftovers.

---

🔐 Privacy & Security

Property	OpenBlue
Outbound telemetry from the extension	None
Analytics / crash reporting from the extension	None
API keys collected by the extension	None
Code sent to a vendor cloud by the extension	None
Network traffic to anything other than your configured backends	None

What does talk to the network?

• LM Studio / Ollama: only when you point OpenBlue at a non-localhost endpoint.
• Model registries (Hugging Face / Ollama registry): only when you explicitly browse or pull a model.
• Claude Code, the Anthropic API, or any other separately-installed third-party tool: subject to their own privacy policies — OpenBlue has no visibility into or control over their network behavior.

TL;DR: if you don't explicitly state a cloud endpoint or web search, OpenBlue is fully airgap-safe.

---

🛠️ Troubleshooting

The 🐻‍❄️ Models don't appear

1. Click the bear → Check Status
2. In the output channel, look for [lmStudioManager] or [ollamaManager] errors
3. Verify the backend is actually listening:

   curl http://localhost:1234/v1/models     # LM Studio
   curl http://localhost:11434/api/tags     # Ollama

4. If both return JSON but the bear is still gray, Reload Window

Claude Code keeps using the cloud, not my local model

1. Open the OpenBlue menu → Toggle Claude Code Routing → ensure it's ON
2. Make sure each role is mapped to a real, currently-available model
3. Restart Claude Code (panel or CLI) — env vars are read at startup
4. Run "OpenBlue: Repair Claude Code ServiceWorker" if the panel is stuck

Inline completions never appear

1. Check the setting openblue.inlineCompletion.enabled is true
2. Confirm the chosen model is loaded (LM Studio) or pullable (Ollama)
3. Try increasing debounceMs if your network/disk is slow
4. Look in the output channel for [inlineCompletionProvider] errors

Out-of-memory / model fails to load

• Lower the quant level (Q5 → Q4 → Q3)
• Pick an MoE alternative
• Reduce context window in LM Studio / Ollama config
• Leave at least 20-30% of VRAM free for KV cache & overhead

Tool calls aren't working

• Not every local model supports tool calling. Pick one tagged agentic / tools / function-calling.
• For MCP servers, verify your mcp.json is valid via "OpenBlue: Configure MCP Servers".
• Toggle web search off if you don't need it — fewer tools = better tool-selection accuracy on smaller models.

---

❓ FAQ

Is OpenBlue free to use?

The standalone edition available on GitHub is free for personal, academic, and evaluation purposes. It runs for 90 days (3 months) from first installation on each device, after which all features are disabled. Reinstalling or resetting the system clock does not extend or restart the trial. For continued use beyond 3 months, a commercial license is required — contact antonfernperez@gmail.com.

Does OpenBlue send my code anywhere?

No. The extension itself has zero outbound telemetry. Your code only goes to the backends you configure (typically localhost). If you choose to use Claude Code alongside OpenBlue, that's a separate Anthropic product with its own privacy policy.

Which is better, LM Studio or Ollama?

Both are excellent. LM Studio has a polished GUI and easier GGUF discovery; Ollama is more script-friendly and has a snappier CLI. OpenBlue treats them as equal first-class citizens — you can even run both simultaneously and route different models through different backends.

Can I use this on a remote server?

Yes. Point openblue.lmStudio.host or openblue.ollama.host at any reachable URL — http://10.0.0.42:1234, https://gpu.lab.internal:11434, etc. SSH tunnels work too.

Why is response quality lower than Claude / GPT-4?

You're running on local hardware — a 14B Q4 model will not match a 1T+ frontier model. Tips:

• Use the largest model your hardware can run comfortably
• Prefer MoE architectures (GLM-4.6, Qwen-MoE, Kimi-K2)
• Keep context short and specific
• Use the right mode (agent for multi-step, ask for single-shot)

Does it work offline?

Yes — once you've installed the extension and pulled the models you want, you can disconnect entirely. The extension makes zero network calls of its own.

What happens if I uninstall?

All configuration OpenBlue wrote (Claude Code env vars, routing files) is cleaned up automatically. Your models stay where they are — they belong to LM Studio / Ollama, not to OpenBlue.

Can I contribute?

This repository is currently a read-only distribution channel. The project is developed and maintained by SIRGPrice. Issues are welcome for bug reports and feature requests. For source-level collaboration, please open an issue or contact the maintainer directly.

---

📄 License

OpenBlue is distributed under the OpenBlue Personal & Research Use License (OB-PRL) v1.0.

In short:

• ✅ Personal use on your own devices
• ✅ Academic research and teaching
• ✅ Studying behavior and publishing observations (with attribution)
• ❌ Commercial use without a separate license
• ❌ Redistribution outside this repository
• ❌ Reverse engineering
• ❌ Using OpenBlue to build a competing product

Read the full license here → LICENSE

For commercial licensing, contact antonfernperez@gmail.com.

---

📋 Third-Party Notices

OpenBlue is an independent product maintained by SIRGPrice and is not affiliated with, endorsed by, or sponsored by any of the third-party products it interoperates with.

Product	Owner	Relationship
Claude Code / Anthropic API	Anthropic, PBC	OpenBlue writes documented config (env vars + JSON files) that Claude Code reads. Not bundled.
LM Studio	LM Studio / Element Labs, Inc.	OpenBlue talks to LM Studio's local HTTP API. Not bundled.
Ollama	Ollama, Inc.	OpenBlue talks to Ollama's local HTTP API. Not bundled.
Visual Studio Code	Microsoft Corporation	OpenBlue is a VS Code extension. Not bundled.
GitHub Copilot	Microsoft Corporation	OpenBlue registers local models so Copilot Chat can list them. Not bundled.

Trademarks. "Claude", "Claude Code", "Sonnet", "Haiku", "Opus", and "Anthropic" are trademarks of Anthropic, PBC. "LM Studio" is a trademark of LM Studio / Element Labs, Inc. "Ollama" is a trademark of Ollama, Inc. "Visual Studio Code", "VS Code", and "GitHub Copilot" are trademarks of Microsoft Corporation. All other names, logos, and brands are the property of their respective owners. Their use in this README is strictly for compatibility identification.

See also NOTICE.md for the long-form notice.

---

Made with 💙 by engineers

---

_{OpenBlue is not affiliated with, endorsed by, or sponsored by Anthropic, LM Studio, Ollama, Microsoft, or GitHub.
All third-party trademarks are the property of their respective owners.}