Lamprey is the transmission for the open-source LLM engine of your choice. It's a desktop coding IDE that grafts Claude Desktop-quality UX onto a Codex-style developer toolset: streaming markdown, reasoning blocks, skills, MCP servers, autonomous loops, sub-agent orchestration, and session memory welded directly onto a file tree, multi-tab Chromium browser, git diff review, integrated terminal, Brave search engine, and side-thread conversations. Plug in DeepSeek, Gemma, Qwen, GLM, or anything on OpenRouter. When the next breakthrough model drops, swap the key and the harness will adapt. Everything stays local, everything persists in SQLite, and API keys never leave the OS keychain. No Token Overlords watching your prompts for the next chance to roach your bank account.
The target user is the developer who looked at Claude Code and Codex and said "I want exactly this, but I'm not paying predatorial prices." Lamprey is a bring-your-own-keys alternative to the two most capable agentic coding tools on the market, built for people who want the power without the leash. It's ready for next-gen models out of the box, as the harness evolves naturally with the ecosystem. 100% vibe-coded over nearly 300 sessions in Claude Code and Codex using WhisprFlow.
| Platform | Format | Link |
|---|---|---|
| Windows x64 | Installer | Lamprey-x64.exe |
| Windows x64 | Portable ZIP | Lamprey-x64.zip |
| macOS Apple Silicon | DMG | Lamprey-arm64.dmg |
| Linux x64 | AppImage | Lamprey-x86_64.AppImage |
macOS note: The DMG is unsigned. On first launch, right-click the app → Open → Open to bypass Gatekeeper. Linux note:
chmod +x Lamprey-x64.AppImagethen run it. All releases: github.com/USS-Parks/lamprey/releases
New in v0.15.6: Retired DeepSeek model cleanup — the deprecated deepseek-chat, deepseek-reasoner, deepseek-v3, and deepseek-r1 API endpoints are removed from the catalog and silently remapped to their live V4 equivalents (deepseek-v4-flash / deepseek-v4-pro). Users with stale model selections get seamless continuity instead of a 400 error.
- Download your platform's installer above and run it.
- Get a key.
- DeepSeek: platform.deepseek.com
- Gemma: (https://openrouter.ai/google/gemma-4-31b-it:free#api) Qwen: GLM: (https://z.ai/manage-apikey/apikey-list)
- Paste your key in the first-run modal. It's encrypted with the OS keychain via Electron
safeStorage. - Type something. Let's go.
- Multi-provider chat — pick a model per task. Cheap for boilerplate, smart for hard bugs.
- Codex-style developer panes — file tree (
Ctrl+P), multi-tab browser (Ctrl+T), git diff review with "Fix this" per-hunk seeding (Ctrl+Shift+G), shell terminal (Ctrl+`), side-thread chat. - Deep Research — research-shaped turns fan out across search providers, corroborate claims by independent domain, and kill the report if they detect fabricated citations.
/research <q>forces it; coding turns are never escalated. - Snip — an in-process token filter (same idea as rtk) that strips noisy shell output down to signal before it hits the model context. ~120 built-in YAML filters, hot-reloadable, extensible.
- Skills + MCP — drop a
.mdin your skills directory and it's part of the system prompt. MCP servers via SSE + stdio with Google OAuth support out of the box. - Loops — set a task on a fixed interval (
/loop 5m <task>), let the model pace itself (/loop <task>), or hand it a mission and a backlog (/loop --auto <mission>) and walk away. The model enqueues work, records outcomes, and self-terminates when the mission is complete. Hard ceilings on iterations, wall-clock, and token budget keep it from running away. Off by default — flip one toggle in Settings to unlock. - Sub-agents — the model can fan out parallel sub-agents via
multi_agent_runwhen the task calls for it. You don't configure this; the model decides when to orchestrate and when to stay single-threaded. - Plan mode —
Shift+Tabblocks mutating tools while read-only tools keep working. Approve, reject, or edit the plan in-place. - Worktrees, forking, hooks, cron automations, AGENTS.md injection, conversation seeding, reasoning trace viewer, optional SQLCipher encryption — see DEVLOG.md for the full build history.
git clone https://github.com/USS-Parks/lamprey
cd lamprey
npm install
npm run devDistributables: npm run build:win / npm run build:linux / npm run build:mac.
Requirements: Node.js 22+, npm 10+, git.
Renderer (React 19 + Zustand)
Sidebar | Chat | Right panel (Tools / Artifacts / Home)
|
| window.api (typed contextBridge)
v
Main process (Node.js)
Provider registry -> DeepSeek / Google / DashScope / OpenRouter
MCP manager (SSE + stdio + OAuth)
better-sqlite3 (WAL, foreign keys)
Browser manager (WebContentsView per tab)
Git runner + review/worktree IPC
Skill loader (chokidar hot reload)
Keychain (safeStorage)
Renderer is sandboxed (contextIsolation, no nodeIntegration). Keys and OAuth tokens never cross the IPC boundary. Artifacts and browser tabs run in isolated Chromium processes.
- API keys encrypted via Electron
safeStorage(OS keychain). - Renderer: sandbox + context isolation + no node integration.
- Artifact sandbox blocks all outbound network and runs in its own process.
- No telemetry. No phone-home. Run it air-gapped if you want.
See CONTRIBUTING.md. Issues and PRs welcome at github.com/USS-Parks/lamprey/issues.
MIT — see LICENSE.
