Sevalla — Claude Code MCP server: deploy and logs
Wires your agent into the Sevalla PaaS: deploys, logs, metrics, env vars and databases without leaving the chat. Useful when the app already runs on Sevalla.
claude mcp add --transport http mcp https://mcp.sevalla.com/mcpWhat Sevalla does
Nothing to install locally. The client talks to a hosted endpoint at https://mcp.sevalla.com/mcp, and sign-in happens through OAuth in your browser with your Sevalla account. No API key goes into the config file.
The agent gets 27 tools. Twenty-two cover the routine work: listing services, triggering and rolling back deployments, reading runtime and access logs, pulling metrics, managing environment variables, scaling processes, adding domains, creating applications, static sites and databases. The other five (sevalla_api_search, sevalla_api_read, sevalla_api_create, sevalla_api_update, sevalla_api_delete) reach the rest of the API by operation ID. The server picks the HTTP method and path rather than the model, and each of those five only runs operations of its own class.
Environment variable values stay hidden: tools show keys and their scoping, and can set or delete them, but never hand a value back. Deleting whole services, managing API keys and operations that mint a fresh secret are all reachable. Destructive tools carry annotations, so the client asks before each call. What the agent can touch is bounded by the Sevalla API key behind the connection — a read-only key keeps it read-only. Worth knowing: commands run inside a process and application logs can still expose secrets.
What it helps with
- read logs from a failed deploy without opening the dashboard
- roll a release back to the previous build
- set an environment variable and restart the process
- spin up a database for a new service
- check app metrics after a traffic spike
Example prompts for the agent
- List my Sevalla services and the status of the latest deployment
- Pull the logs from the failed deployment and find the error
- Set
LOG_LEVEL=debugand redeploy the app
How to install Sevalla
Platform CLI: remote server
claude mcp add --transport http mcp https://mcp.sevalla.com/mcpPlatform CLI: remote server
codex mcp add sevalla --url https://mcp.sevalla.com/mcpConfig file
{
"mcpServers": {
"sevalla": {
"url": "https://mcp.sevalla.com/mcp"
}
}
}Config file
{
"servers": {
"sevalla": {
"type": "http",
"url": "https://mcp.sevalla.com/mcp"
}
}
}Platform CLI: remote server
gemini mcp add -t http sevalla https://mcp.sevalla.com/mcpStep by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Sevalla
How do I install Sevalla in Claude Code?
Run claude mcp add --transport http sevalla https://mcp.sevalla.com/mcp, then type /mcp inside Claude and pick Authenticate to finish the OAuth flow.
What does the Sevalla MCP server do?
It exposes the Sevalla PaaS API to your agent: deployments and rollbacks, logs, metrics, environment variables, scaling, domains and service creation.
Do I need a Sevalla API key?
Not for connecting — OAuth handles that. Permissions still come from the Sevalla API key behind the connection, so a read-only key leaves the agent read-only.
