Official API and MCP¶
Coolify already provides an official REST API and MCP server. ChatCoolify adds Python types, a CLI, and local safety gates around those official interfaces; it does not create a private server protocol.
REST API¶
Common resources include:
- Teams, members, projects, and environments;
- applications, databases, and services;
- servers, destinations, and deployments;
- deployment logs, backups, notifications, and shared variables.
Protected requests use a team-scoped Bearer token:
MCP¶
MCP uses Streamable HTTP. It remains subject to the same instance API switch, team boundaries, token permissions, IP allowlist, and sensitive-value redaction rules.
Permission map¶
| Permission | Use case |
|---|---|
read | AI inventory, status inspection, log summaries |
deploy | CI or controlled AI lifecycle actions |
write | Create or modify projects, applications, and databases |
read:sensitive | Only when supported sensitive reads are truly required |
root | Instance administration; not for normal AI automation |
Enablement order¶
- An administrator enables instance API Access under Settings → Advanced.
- Configure an IP allowlist when callers use fixed source addresses.
- Enable MCP for the intended team.
- Create separate least-privilege, expiring tokens for each AI or CI integration.
- Store tokens in protected ChatEnv or a secret manager.
- Start with health and inventory reads before allowing
deployor short-livedwriteaccess.
Warning
Do not place tokens in prompts, command arguments, repositories, docs, issues, pull requests, or logs.