Working with AI assistants

Deplomo works with Claude Code, Cursor, Codex, Gemini Antigravity and any MCP-compatible agent through the MCP server built into the CLI. An agent can inspect projects and services, deploy source code, read logs and builds, manage variables and domains, and perform lifecycle operations.

Run deplomo login. MCP uses the same saved token.

Connect your agent

Run this in the repository root:


deplomo setup agent

This command writes MCP settings and the use-deplomo guidance skill for recognized agents.

Test the connection

Ask the agent to show the current account and project list, then identify the target service by slug or ID.

Example requests

“Check the latest failed build for the api service and report only the cause.”
“Set NODE_ENV=production as a runtime variable on the web service, then redeploy it.”
“Deploy the current folder as a private service in the demo project.”
“Run deplomo_preflight on this folder, fix any issues, then deploy it and verify that the URL responds.”
“Check the service’s domain and tell me which DNS records it needs. Don’t make any changes yet.”

Access scope and safety

An agent using your full-access CLI token can reach every project available to your account. To limit access, create a read-only, project-limited or spending-capped token with deplomo token create or from Programs → API tokens in the dashboard. Enter a spending cap as an integer number of EUR cents. Secret values are redacted in tool output, but their names remain visible. Deletion, rollback and suspension require confirm: true and should be shown as destructive operations in the client.

Always explicitly identify the project, environment and service. Before confirming destructive operations, check the ID and the effects on volumes, domains and users.

For each agent’s configuration files, global installation and the tool list, see the complete MCP and agents guide.