A Claude Code skill for deploying, managing, and monitoring services on Dokploy via MCP tools.
This skill teaches Claude Code how to interact with your Dokploy instance efficiently — deploy apps, read logs, manage domains, update environment variables, and inspect container status — all through the Dokploy MCP server, never via direct HTTP calls.
npx skills add ppezzull/use-dokploy-mcpOr install only the specific skill:
npx skills add ppezzull/use-dokploy-mcp --skill run-dokploy-mcpCopy the skill into your project:
# From your project root
cp -r .claude/skills/ your-project/.claude/skills/Or add as a git submodule:
git submodule add https://github.com/ppezzull/use-dockploy-mcp.git .claude/skills/run-dokploy-mcp- Dokploy MCP server configured in your Claude Code settings (
.claude/settings.jsonor global):
{
"mcpServers": {
"dokploy": {
"command": "npx",
"args": ["-y", "@dokploy/mcp"],
"env": {
"DOKPLOY_URL": "https://your-dokploy-instance.com",
"DOKPLOY_ACCESS_TOKEN": "your-api-token"
}
}
}
}- Python 3 (for the
extract.pyhelper — handles the 140K+project-allresponse)
| File | Purpose |
|---|---|
SKILL.md |
Agent-facing instructions — auto-loads when deploying or managing services |
extract.py |
Python helper that parses project-all output into compact name+ID listings |
project-all returns 140K+ characters on most instances. Reading it directly blows the context window. The skill provides extract.py to filter:
python3 extract.py <project-all-output-file> my-project# Deploy
application-deploy(applicationId: "...")
compose-deploy(composeId: "...")
# Logs
application-readLogs(applicationId: "...", tail: 50, search: "error")
# Update env vars (newline-separated, not JSON)
application-saveEnvironment(applicationId: "...", env: "KEY=val\nKEY2=val2", ...)
# Domains
domain-byApplicationId(applicationId: "...")
- Never read
project-alloutput directly *-searchtools are unreliable — fall back toproject-all+ filterproject-updaterequiresnameeven if only changingdescription- Compose logs need Docker
containerId, not DokploycomposeId
MIT