Overview
Coolify MCP exposes tools, resources, and prompts for the team associated with the API token. MCP lists only the tools that the token's permissions allow. coolify_help shows these tools and step-by-step recipes, for example topic=deploy_app_from_git.
Start with these tools
| Tool | Use |
|---|---|
coolify_help | Find a tool by intent. |
get_infrastructure_overview | Get counts and health hints for the team. |
search_resources | Find a resource by name, UUID, or domain when its type is unknown. |
list_unhealthy_resources | Find resources that appear unhealthy or down. |
Capability pages
| Page | Covers |
|---|---|
| Teams | Current team, team members, notifications, and audit events. |
| Servers | Servers, metrics, proxy, Docker cleanup, destinations, private keys, registries, and cloud providers. |
| Projects | Projects, environments, resources, search, and tags. |
| Applications | Create, change, and delete applications and preview deployments. |
| Databases | Create, change, import, and back up standalone databases. |
| Services | Create, change, and delete services and their components. |
| Deployments | Deployment history, deploy, cancel, rollback, start, stop, and restart operations. |
| Logs | Container logs and deployment log summaries. |
| Environment and storage | Environment variables, shared variables, storage mounts, and volume backups. |
| Integrations | Git sources, repositories, branches, and S3 storages. |
| Automation | Scheduled tasks, executions, health hints, and unhealthy resources. |
| Prompts and resources | Read-only MCP resources and guided prompts. |
| Pagination | Page arguments and continuation responses. |
Permissions
| Permission | Tools |
|---|---|
read | list_* and get_* tools, and read actions of manage_* tools. |
read:sensitive | Logs, environment values, command bodies, and other sensitive fields. |
deploy | deploy, control, cancel_deployment, rollback_application, and actions that run or delete workloads. |
write | create_* and manage_* tools. They use the same REST API routes, validation, and permissions as the API. |
Destructive actions need confirm=true. Your client must ask you before it sends it.
Most list tools return paginated data. Read Pagination before iterating through large results.