MCP Catalogs
Home

studio

by decocms·371·Score 50

An open-source control plane for AI agents, integrating MCP tools with tracking and organization.

ai-llmdeveloper-toolsproductivity
46
Forks
114
Open issues
this month
Last commit
2d ago
Indexed

Overview

Deco CMS provides a comprehensive management platform for AI agents, enabling users to hire specialized agents, connect MCP tools, and organize them into functional projects. The platform offers a web interface for setting up connections with one-click OAuth, tracking costs and performance, and creating adaptive UIs based on project needs. It supports both local and cloud deployments with team management features, making it suitable for both individual developers and organizations.

Try asking AI

After installing, here are 6 things you can ask your AI assistant:

you:Team-based AI agent management with shared tools and cost tracking
you:Building custom AI applications by composing specialized agents with MCP tools
you:Monitoring and governing AI agent performance and API usage across projects
you:What is the difference between an Agent and a Connection in Deco CMS?
you:How does Deco CMS handle authentication and access control?
you:Can I self-host Deco CMS?

When to choose this

Choose Deco CMS when you need a unified control plane for managing multiple AI agents and MCP tools with project-based organization, cost tracking, and adaptive UIs.

When NOT to choose this

Avoid Deco CMS if you need simple point-to-point MCP integration without the overhead of a control plane, or if you're building revenue-generating SaaS (requires commercial license).

Tools this server exposes

1 tool extracted from the README
  • CONNECTION_CREATECONNECTION_CREATE(name: string, connection: {type: 'HTTP'|'SSE'|'WebSocket', url: string, token?: string}): {id: string, scope: 'workspace'|'project'}

    Create a new MCP connection

Note: Only one explicit tool definition found in the README. The server likely exposes more tools through MCP connections (GitHub, Slack, Postgres, etc.), but these aren't directly documented as tools in the README itself. The 'defineTool' patter

Comparable tools

nexusmcp-server-templatenode-mcp-servermcp-github

Installation

Installation

Quick Start

bunx decocms

From Source

git clone https://github.com/decocms/studio.git
bun install
bun run dev

The server runs at http://localhost:3000 with both client and API server.

Docker

# With embedded PostgreSQL
docker compose -f deploy/docker-compose.yml up

# With external PostgreSQL
docker compose -f deploy/docker-compose.postgres.yml up

Claude Desktop Integration

Add this to your Claude Desktop config.json:

{
  "mcpServers": {
    "deco": {
      "command": "bun",
      "args": ["x", "decocms"]
    }
  }
}

FAQ

What is the difference between an Agent and a Connection in Deco CMS?
An Agent is an AI entity with specific skills and prompts that can be hired and composed. A Connection provides access to tools and services via MCP, which Agents can use to perform tasks.
How does Deco CMS handle authentication and access control?
Deco CMS uses Better Auth for OAuth 2.1 authentication and API keys, with role-based access control (RBAC) at both workspace and project levels.
Can I self-host Deco CMS?
Yes, Deco CMS can be self-hosted using Docker, Kubernetes, or directly with Bun/Node. It runs on your infrastructure with no vendor lock-in.

Compare studio with

GitHub →

Last updated · Auto-generated from public README + GitHub signals.