Supabase MCP

Supabase MCP for Claude Code

Safety first (non-negotiable)

• Do NOT connect MCP to production data. Use a dev project/branch and non-sensitive data.
• Prefer project scoping so Claude cannot access other Supabase projects.
• If connecting to real data is unavoidable, use a read-only mode when available.

(These are recommended mitigations due to LLM prompt-injection risk and data exposure concerns.)

Current project setup

This project is connected to Supabase MCP:

• Project ref: iyyqdoahxhmdumojnqip
• Status: Connected (project-scoped)

Setup (reference)

Option A — CLI (recommended)

Use Supabase remote MCP URL with project_ref (project-scoped):

# Add server (project scope) - replace PROJECT_REF:
claude mcp add --scope project --transport http supabase "https://mcp.supabase.com/mcp?project_ref=PROJECT_REF"

Then authenticate:

1. Run claude /mcp
2. Select supabase → Authenticate
3. Complete browser login (Supabase uses dynamic client registration)

Option B — .mcp.json

Create .mcp.json at repo root: See: templates/mcp.supabase.project.json

CI / headless authentication (optional)

If browser OAuth is not possible (CI), configure an access token header:

• Use Authorization: Bearer ${SUPABASE_ACCESS_TOKEN} and project_ref env vars.

How to use (prompting patterns)

When operating with Supabase MCP tools:

1. Ask for schema introspection first (tables, columns, RLS status)
2. Propose a migration plan before executing write operations
3. If asked to run SQL:
   • Generate SQL
   • Explain it
   • Then run it (only after confirmation if the action is destructive)

Available MCP tools

• list_tables - List all tables in the database
• get_table_schema - Get schema for a specific table
• execute_sql - Execute SQL queries
• list_storage_buckets - List storage buckets
• create_migration - Create a new migration

Output expectations

• Always provide the SQL/migration diff in the answer.
• Always mention what MCP tools were used and what changed.

DeepRead Database Schema (reference)

papers table

CREATE TABLE papers (
  id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
  title TEXT NOT NULL,
  filename TEXT NOT NULL,
  file_path TEXT NOT NULL,
  created_at TIMESTAMPTZ DEFAULT NOW(),
  updated_at TIMESTAMPTZ DEFAULT NOW()
);

paper_pages table

CREATE TABLE paper_pages (
  id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
  paper_id UUID REFERENCES papers(id) ON DELETE CASCADE,
  page_number INTEGER NOT NULL,
  text_content TEXT,
  text_items JSONB,
  width FLOAT,
  height FLOAT,
  has_text BOOLEAN DEFAULT true
);

highlights table

CREATE TABLE highlights (
  id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
  paper_id UUID REFERENCES papers(id) ON DELETE CASCADE,
  page_number INTEGER NOT NULL,
  start_offset INTEGER NOT NULL,
  end_offset INTEGER NOT NULL,
  color TEXT DEFAULT 'yellow',
  note TEXT,
  created_at TIMESTAMPTZ DEFAULT NOW()
);

chat_messages table

CREATE TABLE chat_messages (
  id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
  paper_id UUID REFERENCES papers(id) ON DELETE CASCADE,
  role TEXT NOT NULL CHECK (role IN ('user', 'assistant')),
  content TEXT NOT NULL,
  citations JSONB,
  created_at TIMESTAMPTZ DEFAULT NOW()
);