Your AI harness — local-first memory, skills, tools, and workflows you own and extend.
Not another chatbot. Vodou sits under the places you already work — web AIs, IDEs, CLIs, desktop MCP clients, messaging, and its own Console — with one durable brain on your machine, then arms (MCP tools), playbooks (skills), and scheduling — all expandable by you.
Bring your models. Bring your servers. Bring your workflows. Vodou is the harness that holds them together — and follows you from ChatGPT and Claude.ai to Cursor, Claude Code, and Codex without starting from zero.
vodou.ai · App · Chrome Bridge · Open source · Blog
vodou-bridge-console-memory.mp4
| Layer | What it is | You can expand it |
|---|---|---|
| Memory | Facts, prefs, docs, imports — local DB on your machine | Import sources, vaults, pins, leak policy, what follows you into which surface |
| Skills | Guided workflows with stopping points — expert playbooks, not one-shot prompts | Add, edit, import, and share skills; wire them to the tools you care about |
| MCP tools | Real actions: mail, calendar, browser, tickets, files, APIs… | Connect any MCP server (stdio, HTTP, …). Your catalog, not a fixed vendor list |
| Scripts & jobs | Longer-running work with status and control | Register your own scripts and automation |
| Surfaces | Bridge (web AIs), IDE/CLI hooks, Console, messaging, OpenAI-compatible API, MCP host | Attach the clients you use; scope what each one can see |
| Scheduler / loops | Heartbeats, reminders, things Vodou notices for you | Turn lanes on/off; decide what runs and where it reports |
Everything is meant to be yours to customize: which model answers, which tools are allowed, which vault an agent may search, which skill fires, what never leaves the machine.
Memory that follows you
Press Ctrl+B in ChatGPT, Claude, Gemini, and other supported sites — relevant context lands in their composer. The same brain reaches Cursor, Claude Code, and other coding agents through hooks and rules files. Import ChatGPT / Claude / Obsidian history; pin and correct; keep provenance so your facts outrank throwaway imports.
Skills when a workflow matters
Say what you want done and get a visible plan — run once, edit, save as a reusable skill, or schedule it. Mid-run decisions stop and ask you. Skills are files you can read, change, and grow.
MCP when you need arms
Connect the apps and servers you already trust. Vodou routes natural language to tools and can run them in parallel. Add a new MCP server → new capability. No waiting on our roadmap for your stack.
A host, not only a client
Attach Claude Desktop, Cursor, VS Code, Windsurf, Zed, or your own scripts as MCP clients — each with identity, scope, audit trail, and a kill switch. Or point anything OpenAI-compatible at your local gateway and get your context for free.
Messaging & schedule
Text the same brain from Telegram, Slack, Discord, and more. Heartbeats and automations so work continues when you’re not staring at a chat.
Receipts & honesty
Every turn can leave a receipt (memories · tools · skills). Conflicts surface for you to settle. Graders say unknown when there’s no evidence — never a fake green light.
Your keys, or local
Frontier APIs, or models that never leave your box. Conversation content and memory stay on-device by default.
Web AIs · IDEs · CLIs · Console · messaging · MCP clients
│ Bridge · hooks · channels · MCP · API
▼
┌──────────────────────────────────────────────────────┐
│ Vodou on your machine (the harness) │
│ │
│ Memory ←→ Skills ←→ MCP tools ←→ Scripts │
│ Scheduler · vaults · policy · receipts │
│ Your models · your servers · your rules │
└──────────────────────────────────────────────────────┘
- Capture — chats, docs, pins → local memory.
- Orchestrate — skills + MCP + scripts, under your policy.
- Follow — inject and act on the surfaces you already use.
- Prove — receipts and evidence, not vibes.
macOS / Linux:
curl -fsSL https://raw.githubusercontent.com/VodouAI/OS/main/install-vodou.sh | bashWindows (beta):
irm https://raw.githubusercontent.com/VodouAI/OS/main/install-vodou.ps1 | iexThen finish the wizard at localhost:8765 (account licenses the engine — memory stays local), install Vodou Bridge, pin a fact, open any supported AI, hit Ctrl+B.
Add an MCP server when you want new arms. Drop a skill when you want a playbook. Grow the harness — don’t wait for a single vendor’s memory feature.
Deep docs & architecture: VodouAI/OS.
| Repo | What it is |
|---|---|
| OS | Public open tree — install, Console, Bridge, docs (start here) |
| vodou-core | Engine binaries / core releases |
| vodou-skills-catalog | Skills you can browse and extend |
| lenses-directory | Community page-lenses — PRs welcome |
Models get smarter every quarter. Chat UIs multiply. Built-in “memory” stays trapped in one product. Exporters move transcripts; they don’t give you governed context + tools + workflows under ChatGPT and Claude and your coding agent.
Vodou’s bet: own the harness — memory, skills, MCP, policy — on your machine. Swap models. Add servers. Write skills. Keep the brain.
Public alpha. Built for people who live in multiple AIs and want one stack they can actually extend.
- Product: vodou.ai · app.vodou.ai
- Bridge: Chrome Web Store
- Source: github.com/VodouAI/OS
- Blog: blog.vodou.ai
- Privacy: app.vodou.ai/privacy
Harness · Local-first · Model-agnostic · Expandable by you