Your client just pushed a release. Part of your retainer is making sure it actually works — orders coming in, getting processed, fulfillment closing. But the first order might arrive immediately. Or in 30 minutes. Or an hour. And the client doesn't know either, so they message you. You check. Nothing yet. They message again. You check again. Still nothing. They message a third time.
Each check means opening a terminal, finding the right connection string, spinning up a DB client, running three separate queries, and compiling the results into something readable enough to send back. Twenty minutes, per check, while you're trying to work on something else entirely.
The queries are trivial. Being on call for them all afternoon is not.
What if you could hand some of this off to your AI? Even better — what if you could configure your client's AI so they have the answer without ever disturbing you?
What You're Actually Building
By the end of this, Claude will have direct, controlled access to your Supabase project through a hosted MCP server you configure via MCP Express. You decide exactly which Supabase tools are exposed — nothing more, nothing less.
Examples of what becomes possible:
- "List the tables in the production database and flag any that were created in the last 24 hours."
- "Run this query and tell me if the user_profiles migration applied correctly."
- "What does the current schema look like for the orders table? Generate the TypeScript types."
No terminal context switching. No copy-pasting connection strings. The query goes to Claude, Claude handles it through MCP Express, and you get a structured answer.
Prerequisites
Before you begin, ensure you have:
Estimated time: 5 minutes from signup to first query.
Step 1: Add the Supabase Integration
If you're going through onboarding, you'll be prompted to add your first tool — select Supabase from the list. If you've already onboarded, find Supabase in the tools list inside your dashboard and click on it.
Authenticate via OAuth. MCP Express will request access to the organizations you specify — choose the one containing the projects you want to expose.
After connecting, you'll see the full list of available Supabase tools. This is where cherry-picking matters.
For a typical development workflow, a reasonable starting set:
execute_sql — run queries against your databaselist_tables — schema inspectionlist_migrations / apply_migration — migration managementget_logs — debugging and observabilitygenerate_typescript_types — keep types in sync with schema changesget_advisors — security and performance recommendationsget_project_url / get_publishable_keys — fetch config without touching the dashboard
Leave pause_project, restore_project, delete_branch, and deploy_edge_function disabled unless your workflow explicitly requires them. Fewer tools also means a smaller context window — Claude spends fewer tokens scanning tool definitions it'll never use in that session.
Finally, give your tool a descriptive name so Claude knows when to call it. Something like: "Supabase tools for [Client Name]. Use when asked about the database schema, running queries, or inspecting migrations."
Step 3: Connect Claude to Your MCP Server
You have two options depending on how you use Claude:
The easy way — OAuth (recommended)
In Claude, go to Customize → Connectors and click Add new Connector. Give it a name and set the URL to:
https://api.mcp-express.com/gateway/mcp
Claude will redirect you to MCP Express to complete the OAuth flow. Log in, grant access, and select the MCP server you just configured. No config files, no CLI, no Node.js required.
The manual way — MCP Express CLI
If you prefer the manual config route, we've covered it step by step in our connection guide. It takes under 5 minutes.
Step 4: Start Querying
Open Claude and ask in plain English. Claude uses your configured Supabase tools to interact with the database and return a structured answer.
Start with a simple verification: "List the tables in my database. Use Supabase MCP tools." If the connection is working, Claude will call list_tables and return your schema.
One of the more useful things MCP Express lets you do with Supabase is limit what the agent can actually touch.
The Supabase integration exposes tools that cover the full lifecycle of a Supabase project — from creating one to pausing it. For most workflows, you don't want all of that available in the same session. A content workflow agent querying tables doesn't need pause_project. A schema review agent doesn't need deploy_edge_function.
Use MCP Express's tool cherry-picking to define exactly which tools are included for a given connection. This reduces the blast radius if a prompt goes sideways, and it keeps the LLM context window from filling up with tool definitions it'll never use — a real performance consideration when you're working with 30+ available functions. You're also spending fewer AI tokens, since the model only sees the tools you've explicitly enabled.
Your Supabase Project Is Now Connected
If you already have a Supabase project and an MCP-compatible AI client, you can be connected in under five minutes. Start with a minimal tool set (execute_sql, list_tables, get_logs), verify it works, and expand from there based on what your workflow actually needs.
The full Supabase tool list is available in the MCP Express integration panel once you've connected.
Connect Supabase to Claude →
Further Resources
- Documentation — Every supported integration, configuration option, and code example in one place.
- Contact Us — Got a question before signing up, or want to talk through your setup? Drop us an email.
- Open a Support Ticket — Already inside the app and something's not working? Open a ticket directly from your dashboard.
