Island Mountain - local AI infrastructure
Home Why Local AI Summit Landfall Technology Pricing Solutions FAQ Blog Resources Request Quote
Free Download · MIT License · Windows 10/11

Lamprey

A multi-provider AI coding IDE. Route prompts to DeepSeek, Gemma, Qwen, or OpenRouter. Run a Planner → Coder → Reviewer pipeline. Git review, browser, terminal - all in one window. No telemetry. No subscription. Your API keys, your machine.

Windows 10/11 x64 · Open source on GitHub · Latest release →

Before you run: this build is not yet Authenticode code-signed. Windows SmartScreen will show an “Unrecognized publisher” warning. Click More info, then Run anyway. Verify the SHA-256 hash below before running.

What Lamprey is

Lamprey turns a multi-provider AI backend into a working coding IDE. Think Claude Desktop quality UX combined with Codex-style developer panes, with the freedom to choose your own model and keep every prompt on your own machine.

It is an Electron app. It routes per-model to four providers. It runs a multi-agent pipeline if you want one. It ships full developer tooling around the chat. Everything persists locally in SQLite. No Lamprey cloud. No telemetry. Your prompts go to whichever provider's key you've added - nowhere else.

Free MIT License Open Source No telemetry SQLite local storage Air-gap capable
DeepSeek Google Gemma Alibaba Qwen OpenRouter

What's included

Every feature below runs on your machine. No cloud required.

Multi-agent pipeline

Assign a different model to Planner, Coder, and Reviewer roles. Each stage gets its own model tier - cheap for boilerplate, smart for hard bugs. Chain-of-thought from every agent is persisted and surfaced in the UI.

Codex-style developer panes

File tree with fuzzy quick-open (Ctrl+P), multi-tab Chromium browser, git diff review with a per-hunk "Fix this" button that seeds the diff into the chat, and an integrated xterm.js terminal.

Deep Research

Twelve-prompt pipeline that fans out to DuckDuckGo, Brave, or SerpAPI, deduplicates by domain, corroborates claims across sources, and rejects fabricated citations before the report reaches you.

Snip - token filter

~120 built-in YAML filters strip verbose shell output down to signal before it reaches the model. Git, npm, docker, kubectl, terraform, and more. Hot-reloads from your userData folder in ~1 second. Tracks tokens saved.

Automations and hooks

Schedule prompts with a cron editor. Fire JavaScript hooks on session start, prompt submit, agent stop, or pre/post tool use. Hooks can block a tool by throwing. Everything runs locally.

Plan mode and permissions

Shift+Tab gates mutating tools while read-only tools keep running. Persistent per-tool allow/deny policies survive restart. Destructive Chrome MCP actions require explicit approval with a 30-second timeout.

OS keychain key storage

API keys are encrypted via Electron safeStorage (the OS keychain). Keys never cross the IPC boundary and never touch disk in plaintext. A yellow banner warns you if safeStorage is unavailable before falling back.

Skills and MCP servers

Hot-reloading markdown system-prompt fragments (drop a .md in userData/skills/). Built-in MCP servers for Gmail, Drive, and Chrome via Playwright. SSE + stdio transports. URL import for skills.

Worktrees

Create and manage git worktrees from inside the app. Each worktree seeds a new conversation tagged to that branch. Sidebar badges distinguish local, cloud, and worktree threads.

Installer vs. portable

NSIS installer (.exe) - Writes to Program Files, creates Start Menu and desktop shortcuts, adds an uninstaller. Best for everyday use.

Download Installer

Portable ZIP (.zip) - Unzip anywhere, run Lamprey.exe directly. No installation needed. Useful for restricted machines or quick evaluation.

Download Portable ZIP

Verify your download

SHA-256 hashes for every release are published alongside the binaries on the GitHub Releases page.

View SHA256SUMS on GitHub

Compute locally once downloaded (PowerShell):

Get-FileHash -Algorithm SHA256 Lamprey-x64.exe

Or use sha256sum -c SHA256SUMS on Linux/macOS.

System requirements

  • Windows 10 or 11, 64-bit
  • Administrator rights for the installer (portable ZIP: not required)
  • ~400 MB free disk space (Electron runtime included)
  • At least one AI provider API key to chat
  • Internet access to reach your chosen AI provider
  • Node.js not required for end users
  • Linux and macOS: buildable from source (GitHub)
  • For local model inference: see Technology

Quick start (four steps)

1
Download and run the installer

Click More info then Run anyway on the SmartScreen prompt. Step through the NSIS wizard - the defaults are fine. Or unzip the portable ZIP anywhere and run Lamprey.exe directly.

2
Get an API key

Easiest start: platform.deepseek.com - sign up, create a key, load $5. Lamprey also accepts Google AI Studio (Gemma), Alibaba DashScope (Qwen), and OpenRouter keys. Or point it at a local model server.

3
Paste your key in the first-run modal

Lamprey encrypts it with safeStorage (the OS keychain). The key never leaves your machine in plaintext.

4
Type something

That's it. Open the + menu to add a file tree, browser tab, terminal, or git review pane. Use Shift+Tab to engage plan mode. Type /research to kick off a deep research run.

What to expect during install

1
"Windows protected your PC" (blue screen)

SmartScreen flags unsigned binaries. Click More info, then Run anyway. We are working on an EV Authenticode certificate - once it ships, this warning disappears on future builds.

2
User Account Control (UAC) prompt

The installer writes into Program Files, so it requests administrator rights. Click Yes. Lamprey itself runs as your normal user after install. The portable ZIP skips UAC entirely.

3
NSIS setup wizard

Step through the wizard and accept the defaults. On the final page, Create a desktop shortcut is checked by default - keep it. Lamprey installs to C:\Users\[You]\AppData\Local\Programs\lamprey\.

4
First launch

Lamprey opens to a first-run modal asking for an API key. Paste your DeepSeek (or other) key, confirm, and the main chat window appears. Add AGENTS.md to your project root to give the model repo-specific context on every send.

What you're installing

Lamprey is a single Electron application. The installer places it in C:\Users\[You]\AppData\Local\Programs\lamprey\ and creates Start Menu and desktop shortcuts. Key facts:

  • No background services. Nothing runs unless you launch Lamprey.
  • No scheduled tasks, no system-tray app, no startup entries.
  • All conversation data, API keys, and settings stay in your user's AppData\Roaming\lamprey\ folder.
  • No telemetry, no crash reports, no usage data sent anywhere.
  • To uninstall: Settings → Apps → Installed apps → Lamprey → Uninstall.

See the Privacy Policy for the full data handling statement.

Summary: Lamprey is a free, open-source Electron desktop IDE for Windows that routes AI chat to DeepSeek, Google Gemma, Alibaba Qwen, or OpenRouter. It runs a configurable multi-agent pipeline, ships Codex-style developer panes (browser, terminal, git review, file tree), and keeps every prompt, key, and conversation on your local machine. No Lamprey cloud. No subscription. MIT license. Two downloads: an NSIS installer and a portable ZIP.

Frequently asked questions

What is Lamprey?

Lamprey is a free, open-source Electron desktop app that combines multi-provider AI chat with a Codex-style coding IDE. It routes prompts to DeepSeek, Google Gemma, Alibaba Qwen, and OpenRouter. Bring your own API keys. MIT license. No telemetry.

Does it cost anything?

No. Lamprey is free and open source under the MIT license. You pay only for the AI provider API calls you make. DeepSeek offers a free tier to start. There is no Lamprey subscription, no seat license, and no cloud service to pay for.

Which AI providers does Lamprey support?

Lamprey routes per-model to DeepSeek (V4 Pro, V4 Flash, V3, R1), Google Gemma via AI Studio, Alibaba Qwen via DashScope, and any model on OpenRouter. You can assign a different model to each agent role - Planner, Coder, Reviewer - in the same pipeline run.

Does Lamprey need an internet connection?

Lamprey itself runs locally with no telemetry or phone-home. It does need to reach whichever AI provider API you configure to send prompts. If you pair it with a local Island Mountain inference server running vLLM or Ollama, you can route to local models and run fully air-gapped.

Why does Windows warn me when I run this?

The installer is not yet Authenticode code-signed. SmartScreen flags any binary from an unrecognized publisher. Click More info, then Run anyway. Verify the SHA-256 hash on this page against the file you downloaded before running - that is the primary trust signal until the EV certificate arrives.

What is the difference between the installer and the portable ZIP?

The NSIS installer (.exe) writes to Program Files, creates Start Menu and desktop shortcuts, and adds an uninstaller. No UAC is needed after first install. The portable ZIP (.zip) unpacks anywhere and runs Lamprey.exe directly - no install, no admin rights required. Both run the same app and the same version.

Can I use Lamprey with a local model instead of a cloud API?

Yes. Point Lamprey's OpenRouter-compatible endpoint at any local server that speaks the OpenAI chat completion API - including vLLM, Ollama, or an Island Mountain Summit inference server. No cloud keys required for local routing. See the Technology page for local inference options.

Can I use Lamprey on Linux or macOS?

The pre-built download targets Windows 10/11 x64. Linux and macOS are buildable from source using npm run build:linux or npm run build:mac. Requirements: Node.js 22+, npm 10+, git. See the GitHub repo for full build instructions.

Where can I file issues or ask questions?

Open an issue on GitHub, email [email protected], or call 1-801-609-1130. Include the version number (shown in Settings → About or the title bar) and a description of what you expected vs. what happened.

Want local inference without cloud API costs?

Pair Lamprey with an Island Mountain Summit server and route to models running entirely on your own hardware. No API bills. No data leaving your network. One-time hardware purchase.

See Summit Servers Request a Quote

Or call directly: 1-801-609-1130