auxilia
keurcien/auxilia
A lightweight web MCP client designed to host your MCP-powered AI assistants. auxilia is designed to only support remote MCP servers.
Instalação
npx convez-ai keurcien/auxilia1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
README
A tradução para o português deste README ainda está na fila.
https://github.com/user-attachments/assets/3236f9da-28c5-44c8-8c2c-4c68199f187c
Why auxilia
Most MCP clients today are desktop apps tied to a single user. auxilia is different:
- 🌐 Web-based — accessible from any browser, no desktop app to install
- 👥 Multi-user — admins register MCP servers once; every teammate can bind them to agents
- 🔌 Remote MCP only — built for server-to-server Streamable HTTP connections, not local stdio
- 🏠 Self-hosted — runs on your infrastructure; credentials and conversation data stay in your database
- 🧩 MCP-native — tools, skills, and resources all flow through MCP, keeping the core tiny
Learn more about auxilia: https://auxilia-docs.vercel.app/
✨ Features
🤖 Agents
- Define agents with a system prompt, avatar (emoji + color) and bound MCP servers
- Subagents — coordinator agents can dispatch work to specialized subagents
- Workspace roles & per-agent permissions —
owner,admin,editor,user, plus private/shared agents - Streaming responses over SSE via the LangGraph SDK, with full LangGraph checkpointing
🔧 Tools & MCP Servers
- Register remote MCP servers at the workspace level — everyone can reuse them
- Per-tool control: always allow, needs approval, or disabled
- Human-in-the-loop approvals inline in chat or Slack
- AES-GCM encryption for API keys at rest, per-user OAuth 2.1 token storage in Redis
- Built-in connectivity probes with automatic token refresh
🧪 Code Sandbox
Turn any agent into a lightweight data-analyst with an isolated Linux environment powered by OpenSandbox:
- Filesystem tools:
ls,read_file,write_file,edit_file,glob,grep - Shell access via
execute - Lazy-spawned containers with a 30-minute TTL, reconnectable across browser refreshes
⚠️ To enable code execution, you first need to deploy Opensandbox (either on a simple VM or Kubernetes), and set it up in auxilia.
💬 Slack Integration
- Start a chat with auxilia and pick an agent
- Tool-approval buttons rendered with Block Kit
- Slack ↔ auxilia users matched by email
📊 Observability
- Langfuse integration — trace every LLM call, tool call, and cost per agent and user
- Request-scoped profiling via
RequestTimer(seeapp/utils/timer.py)
🔐 Authentication
- JWT sessions over HttpOnly cookies
- Google OAuth sign-in (Authlib)
- Personal Access Tokens (PATs) for programmatic access
- Argon2 password hashing
🚀 Quick Start
Prerequisites
- Docker and Docker Compose
- At least one LLM API key (Anthropic, OpenAI, Google, or DeepSeek)
Run with Docker
git clone https://github.com/keurcien/auxilia.git
cd auxilia
cp .env.example .env # add your LLM API keys
make build && make up
Open http://localhost:3000. The first account you create becomes the workspace admin.
Run for development
cp .env.example .env
make dev
make dev starts PostgreSQL, Redis, the FastAPI backend (with migrations applied) and the Next.js frontend in parallel — all with hot reload.
See the Get Started guide for the full walkthrough.
🤝 Integrations
LLM Providers
| Provider | Models |
|---|---|
| Anthropic | Claude Haiku 4.5, Sonnet 4.6, Opus 4.6 |
| OpenAI | GPT-4o-mini |
| Gemini 3 Flash Preview, Gemini 3 Pro Preview | |
| DeepSeek | DeepSeek Chat, DeepSeek Reasoner |
MCP Servers — one-click install
Notion · Linear · GitHub · HubSpot · BigQuery · Slack · Sentry
Plus any custom remote MCP server — just paste a URL and your OAuth credentials or API keys.
Workspace integrations
- Slack — invoke agents, approve tools, stream replies in threads
- Langfuse — LLM + tool-call tracing and cost attribution
- Google OAuth — SSO for your workspace
💡 What you can build
- A CRM agent that queries HubSpot and drafts Slack replies
- A data analyst that runs BigQuery and stores the analysis in a Notion document
- An on-call assistant that reads Sentry issues, searches Linear, and opens PRs via GitHub
- A workspace coordinator that dispatches work to specialized subagents
📚 Documentation
- Get Started — run auxilia locally
- Agents — configuration, permissions, subagents
- MCP Servers — registration, auth, examples
- Tools — per-tool approval rules
- Sandbox — enable code execution for an agent
- Deployment — Docker Compose, Google Cloud Run
- Integrations — Slack, Langfuse
🗺 Roadmap
- Scheduled tasks
- Skills
- Support for more sandboxes (e.g. Daytona)
- Deployment guides
- More SSO providers (Okta, Entra ID)
Have an idea? Open an issue or start a discussion.
🤝 Contributing
Contributions are very welcome! Whether it's a bug report, an agent idea, a new MCP server recipe, or a UI improvement:
- Fork the repo and create a branch (
git checkout -b feature/my-change) - Run
make devand make sure tests pass (cd backend && uv run pytest) - Open a pull request
Please read CLAUDE.md in the root of the repo — it documents the backend's layered architecture (router → service → repository → model) and naming conventions.
📄 License
AGPL-3.0 — free to use, modify, and self-host.