fazer > falar.
We're engineers from São Paulo. We put AI agents to work in production, from customer service on WhatsApp to development infrastructure inside client teams. Most of what we build is open source.
Our product, open source: AI agents that run your customer service. They talk like real people: they transcribe voice notes, read images, reply by text or voice, query a knowledge base, schedule appointments, collect payments, move the sales pipeline and know when to hand off to a human. Self-hosted, with any model: OpenAI, Anthropic, Google, DeepSeek, OpenRouter or local models via an OpenAI-compatible endpoint.
One command takes an empty VPS to an agent answering on WhatsApp, guided by an AI agent in your terminal:
curl -fsSL https://app.fazer.ai/api/agents/install.sh | bash| Repo | What it Does |
|---|---|
| agents | The platform: LangGraph TS runtime, multimodal pipeline (STT, vision, TTS), RAG, operator console, REST API and MCP server. Free edition under Apache 2.0. |
| agents-skills | The AI skills (Claude Code, Codex, Hermes) that install, operate and extend the stack: onboarding, operation and development. |
The Pro edition adds multi-tenant and white-label. Details at app.fazer.ai.
| Service | What you get |
|---|---|
| AI Development Consulting | Claude Code, Codex and Gemini deployed as development infrastructure inside your team. Context engineering, autonomous agents, sandboxing and governance baked in. From 30 days to 1. |
| Fast Testing at Scale | A subscription that handles your AI backlog in isolated, production-safe environments. We validate before you invest in development. Recent runs: 70,000+ people served and 1.2M+ WhatsApp messages processed in a 30-day window. |
| Special AI Projects | End-to-end engineering for custom AI. Modern stack (LangGraph, NVIDIA H100, Kubernetes, Cloudflare Zero Trust). Banking, industrial, events, retail and real estate. |
Tools we maintain and run in production, starting with the WhatsApp + Chatwoot layer that fazer.ai agents runs on.
| Repo | What it Does |
|---|---|
| chatwoot | Our Chatwoot fork: multi-provider WhatsApp (Baileys, Z-API, 360Dialog, Cloud API), native group conversations, internal agent chat, scheduled and recurring messages, full pt-BR support. |
| baileys-api | HTTP REST wrapper around the Baileys WhatsApp library. Connect multiple accounts, send and receive messages, manage sessions. Bun and Elysia. |
| mcp-chatwoot | Model Context Protocol server exposing 129 Chatwoot tools to Claude Desktop, VS Code, and other MCP clients. |
| n8n-nodes-chatwoot | n8n community node with 40+ operations, including the Kanban, scheduled messages and WhatsApp provider features exclusive to our fork. |
| mcp-obsidian | MCP server for Obsidian. Give your AI agents read and write access to your vault. Published on npm. |
| langfuse-proxy | Lightweight Bun and Elysia proxy for OpenAI-compatible APIs with Langfuse telemetry built in. |
- Fork the repository you want to improve.
- Branch off main (
git checkout -b feature/your-thing). - Open a PR with a clear description and test plan.
Our team reviews actively. For larger or strategic contributions, please reach out first via the contact channels below.
Each repository states its own license. fazer.ai agents (Free edition) is Apache 2.0; unless otherwise noted, everything else we release is MIT.
| Website | fazer.ai |
| WhatsApp (Strategic Consulting) | +55 11 5199-9937 |
| contact@fazer.ai | |
| Office | Av. Brigadeiro Faria Lima, 1811, Conj. 115, Jardim Paulistano, São Paulo, SP, 01452-001, Brazil |
fazer > falar.
