The future of agentic voice experiences. Now.
Get started in 2 minutes — install with curl | sh.
→ smartchats.ai · docs
curl -fsSL https://smartchats.ai/install | bashThen:
smartchats setupThat's it. The wizard walks you through your API keys, builds what it needs, and opens http://localhost:3000. No Docker, no Node, no prereqs — the installer bundles the runtime.
Other install paths: npm install -g smartchats-ai (if you already have Node 24+), npx smartchats-ai (no install), git clone + npm link (contributors). Full options at docs/install.
Voice in. Code, charts, web answers, persistent knowledge — out. SmartChats turns conversation into a computing surface that gets smarter every time you use it.
- Voice-native, cross-device. Talk naturally, interrupt, switch devices mid-thought. Same agent, same memory.
- Multi-model routing. GPT-5.5, Claude Opus 4.7, Gemini 3.1 Pro — pick per task or per turn.
- Sandboxed code execution + a knowledge graph that compounds. The agent writes JavaScript that runs in your browser, persists what it learns about you to a local SurrealDB triple store with vector search.
- MCP both directions. Consume external MCP servers; expose SmartChats as one to Claude Desktop or any MCP-aware tool.
- Self-hostable. The entire stack runs on your laptop with one command. Hosted SaaS optional.
→ Read the docs for architecture, full CLI reference, MCP integration, package guides, and contribution workflow.
🟢 Production-ready · Open-core release · 2026 Q2
The hosted SaaS at smartchats.ai runs the same stack you can self-host here. Only multi-tenant cloud orchestration (billing back-end, hosted database, infrastructure) stays private.
smartchats setup guided first-run
smartchats start launch the local stack
smartchats stop stop the stack
smartchats status what's running + health
smartchats logs tail per-process logs
smartchats doctor environment health check
smartchats data import / export user data
smartchats login sign in to the hosted SaaS
Full reference: docs/cli.
| Package | What it does |
|---|---|
tivi |
Browser voice interface: VAD, STT, TTS, calibration |
cortex |
Function-calling agent runtime + sandbox executor |
smartchats-database |
Pure SurrealQL query builders + ops layer |
smartchats-backend |
HTTP transport contract, streaming helpers |
smartchats-sessions |
Session export + per-session analyzers + cross-session triage |
llm-service |
Provider-agnostic LLM client (Anthropic / OpenAI / Gemini) |
simi |
Declarative E2E workflow tests |
smartchats-cli (npm: smartchats-ai) |
The smartchats command-line tool |
smartchats-mcp |
MCP server — expose SmartChats to Claude Desktop / any MCP client |
Plus supporting libraries: graph-viz, smartchats-common, smartchats-local-server, smartchats-backend-local, smartchats-cloud-client, smartchats-test.
| Quarter | Milestone |
|---|---|
| Now (Q2 2026) | Open Core + Hosted. MIT-licensed source release. Hosted web app live with managed billing, auth, and infrastructure. |
| Q3 2026 | Integrations + Mobile. Native mobile app launch. Third-party integrations buildout — Gmail, Calendar, X, GitHub, and more — so SmartChats can read, write, and act across the apps where users already live. |
| Q4 2026 | Enterprise. Three offerings: drop-in voice agent SDK (Tivi), drop-in voice + agent runtime (Tivi + Cortex), and full closed-cloud licensing with deployment support for regulated environments. White-label and self-hosted across all tiers. |
A full contribution policy is being prepared. In the meantime, reach out before opening pull requests:
Working on SmartChats itself? See docs/contributing for the npm run dev / bin/devserve workflow.
- 🐛 Issues — bugs, feature requests, questions
- 💬 Discussions — design conversations, show-and-tell
Sheun Aluko, MD, MS — Founder & CEO