Overview

The home for your AI agents

deco Studio is where you build agents, connect them to tools, run them on schedules or events, and see exactly what they’re doing. Self-host or use the cloud. Built on the Model Context Protocol.

Agents

An agent is an AI configured for one specific job: a focused set of tools, clear instructions, and its own thread history. Agents are the working surface of Studio — you spend most of your time inside one. Each agent is also its own MCP endpoint, so external clients like Cursor and Claude Desktop can use it directly.

Connections

Plug in any tool or service — GitHub, Slack, databases, your own API, anything MCP-compatible. Set up credentials once at the org level, then attach the tools you want to each agent. Credentials never leave Studio.

Automations

Run an agent unattended on a schedule or in response to an event. Attached to a specific agent, so the scope of what the run can do is whatever you’ve configured for that agent.

Observability

Every tool call your agents make is logged automatically — who called it, when, what came back, and how long it took. One place to debug, audit, and watch performance.

Extensibility

Studio is open at every layer:

MCP apps — bundle tools with React views in a single package, and Studio will render those views directly inside the agent. Start from the template repo.
Sandboxes — back an agent with a GitHub repository and Studio runs it as an isolated dev environment scoped to that agent (typically the agent’s own MCP server).
Git-native — code-backed agents live in a real repo with branches and PRs, surfaced in the agent’s Git tab. Your version control stays your source of truth.
MCP everywhere — every agent is exposed as its own MCP endpoint, callable from Cursor, Claude Desktop, your own code, or another agent.

Solutions: pre-built agent suites

For teams managing storefronts and content-heavy digital experiences, deco offers packaged agent suites built on top of Studio — the most complete one available today is the Agentic CMS, the evolution of deco.cx. See Solutions for the full list.

What you can do with Studio


Build agents	Compose focused AI workers from your connected tools, instructions, and resources
Connect tools	Add GitHub, Slack, databases, or any MCP server — credentials stay secure
Run hands-free	Trigger agents on schedules or events with automations
See everything	Logs, costs, and traces for every agent action, in one dashboard
Extend with code	Ship MCP apps (tools + UI) and back agents with Git repos in their own sandboxes
Control access	Manage who can use which tools with team roles and permissions
Run on your terms	Self-host on your own infrastructure or use our cloud

Deployment options


Local	`bunx -p decocms deco` — runs entirely on your machine, fully private
Cloud	studio.decocms.com — manage everything from your browser
Team	Shared connections, roles, and cost tracking across your organization
Self-hosted	Deploy to your own infrastructure with Docker or Kubernetes

Get started

Quickstart — Get running in 5 minutes (cloud or local)

Key concepts — Anatomy of the Studio UI, plus how the pieces fit together

Self-hosting — Docker Compose or Kubernetes

The home for your AI agents

Agents

Connections

Automations

Observability

Extensibility

Solutions: pre-built agent suites

What you can do with Studio

Deployment options

Get started

Links