The indie stack in one note: database health, deployments, traffic and errors — read-only by default.
Use case: the typical indie-dev stack — Supabase/Neon database, frontend on Vercel, DNS/workers on Cloudflare — administered and queried from project notes. All four have official MCP servers.
{
"mcpServers": {
"supabase": {
"command": "npx",
"args": ["-y", "@supabase/mcp-server-supabase@latest", "--read-only"],
"env": { "SUPABASE_ACCESS_TOKEN": "sbp_..." }
},
"neon": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://mcp.neon.tech/sse"]
},
"vercel": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://mcp.vercel.com"]
},
"cloudflare": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://observability.mcp.cloudflare.com/sse"]
}
}
}
Supabase: personal token + the recommended
--read-onlyflag. Neon, Vercel and Cloudflare: remote servers with OAuth throughmcp-remote.
On my Supabase project: list the tables with their row counts, and
flag the ones without Row Level Security enabled.
The last 5 deployments of the "marketing-site" project: status,
duration, branch. Does the latest build have warnings?
On the example.com zone: requests and error rate over the last 24h.
Is there an unusual spike, and from which countries?
List my Neon database branches with their sizes. Can the ones created
for previews more than 30 days ago be deleted?
--read-only / minimal-scope tokens: reading covers 90% of the
use cases.