Deploy a private AI workspaceacross your organization.

Connect private MCPs. Configure agents with team context. Give every group the right models and tools—with SSO, permissions, audit logs, and cost controls.

Become a contributor

Workspace

Threads

Find the checkout failures from the last deploy, fix the root cause, and open a PR.

SentryRead 184 checkout errors

PostgresCompared failed order payloads

I found a schema mismatch introduced in the latest checkout deploy. Tracing the failing requests and checking the repo now.

41 tok·

GitHubOpened fix/checkout-schema

Fixed the schema mismatch and opened PR #1421. Error rate is back to 0% over the last 5 minutes.

33 tok·

Ask anything, / for prompts, @ for agents & resources…

Tools

Thinking

Agent

Reliability AgentClaude CodeThinking

Connected tools

Sentryread

Postgresread

GitHubwrite

Output

#1421Open

fix/checkout-schema

This thread

$0.4296k tokens

Runtimes and providers

Claude Code

Codex

Anthropic

OpenAI

Gemini

OpenRouter

MCP

Ollama

The architecture

Your teams get vibecoding powers.
You keep control.

Studio packages the infrastructure behind an internal AI rollout: model routing, MCP authentication, agent configuration, SSO, RBAC, audit logs, and usage accounting.

Start with one team. Standardize approved models, tools, and context. Expand across the organization without copying secrets or rebuilding the platform.

The workspace

One interface for
agents, tools and models.

Package context, tools, and policy into an agent.

Define instructions, add skills and files, grant approved MCP access, choose a model policy, then publish the agent to the organization.

Acme agents

Approved agents available to the organization.

Search agents

verified

Code reviewer

Engineering · 8 tools

shared

Deep research

Strategy · 5 sources

verified

Release manager

Engineering · 6 tools

shared

Commerce operator

Commerce · 14 tools

verified

Frontend QA

Engineering · 9 tools

shared

Incident response

SRE · 7 tools

Package context, tools, and policy into an agent.

Define instructions, add skills and files, grant approved MCP access, choose a model policy, then publish the agent to the organization.

Organization rollout

Start with one team.
Expand without rebuilding.

Organization and local runtimes

Use the organization router—or your own coding agent.

Shared chat runs through the organization AI router, backed by OpenRouter or direct providers. For coding work, engineers can link their own Claude Code or Codex session and use the subscription already authenticated on their machine.

Runtime for this thread

Choose per conversation or agent

Organization AI router

OpenRouter · direct providers

shared

Local coding agent

Claude Code · Codex

your machine

Local authentication and subscription stay on your machine.

Private MCP access

Share MCP capabilities—not credentials.

Connect GitHub, Slack, Postgres, Sentry, or an internal MCP once. Store tokens in the encrypted vault, then grant tool-level access by organization, role, or agent.

Engineering MCP

Shared with 18 teammates

GitHub · write

Sentry · read

Postgres · read

Slack · write

AES-256-GCM token vault encrypted

Usage and audit

Account for every model and tool call.

Trace the user, agent, model, tools, latency, errors, tokens, and cost for every thread. Break usage down by agent, connection, organization, or teammate.

AI spend · 30 days

$1,842.20

−18.4%

Claude Code$824.10

OpenRouter$611.42

Decopilot$406.68

Organization controls

Roll out with the controls your security team expects.

Use SSO, role-based access, organization isolation, tool-scoped API keys, and audit logs. Run on your infrastructure with your identity provider, model keys, and data stores.

Organization policy

Applied to all members and agents

active

Single sign-onSAML / OIDC

Role-based access4 roles

Tool permissionsscoped

Audit exportenabled

Deployment

Run Studio on your infrastructure.

Local, Docker, or Kubernetes. Bring your own database, keys, and storage.

View repository Deployment docs

~/deco/studio

$ bunx decostudio

starting deco Studio…

✓ embedded PostgreSQL ready

✓ encrypted token vault ready

✓ MCP gateway listening

Studio is running at http://localhost:3000

Local

One command and embedded Postgres.

Docker

Published image and Compose files.

Kubernetes

Helm chart for production rollout.

Enterprise

SSO, RBAC, audit, and your keys.

Get started

Contribute to deco Studio.

Studio is open source and self-hosted. Contribute on GitHub, report issues, or deploy on your own infrastructure.

Become a contributor

Open source · self-hosted · no credit card

Legal

Integrations

Google Oauth

Deploy a private AI workspaceacross your organization.

Your teams get vibecoding powers.You keep control.

One interface foragents, tools and models.

Start with one team.Expand without rebuilding.

Use the organization router—or your own coding agent.

Share MCP capabilities—not credentials.

Account for every model and tool call.

Roll out with the controls your security team expects.

Run Studio on your infrastructure.

Contribute to deco Studio.

Service

Software

Company

Social

Legal

Integrations

Your teams get vibecoding powers.
You keep control.

One interface for
agents, tools and models.

Start with one team.
Expand without rebuilding.