Supabase
Supabase prompts now let AI agents inspect production projects for health, security, performance and resource use. Checks can run on demand or to a schedule inside Claude, Codex or Cursor.
Agents can query project logs directly and report what they find. The developer still decides whether any resulting change ships.
The update, shared in a LinkedIn post, also adds controls over which people, agents and projects can connect.
Supabase Prompts Add Production Checks With Guardrails
Companies can control access through their identity provider. Tokens can be restricted to selected projects rather than giving an agent access to every Supabase project in an organisation.
The agent must also ask for approval before creating a paid project or running SQL that could delete data. These approval points sit between the agent’s proposed action and the production system.
| Area | What the agent can do | Control |
|---|---|---|
| Project operations | Check health and resource use | Run on demand or to a schedule |
| Security and performance | Inspect the project and report findings | Developer decides what ships |
| Logs | Query production logs directly | Tokens can be limited to selected projects |
| Higher-risk actions | Create paid projects or run potentially destructive SQL | Agent requests approval first |
The identity and token controls follow Supabase making enterprise-managed auth for its MCP server generally available. Together, these controls define who may connect and which projects an agent may reach.
Data Streaming Moves Heavy Analytics Off the App Database
Supabase has also put data streaming into public alpha. Project data can be streamed to BigQuery, ClickHouse, DuckLake or Snowflake.
This separates heavy analytics work from the application database. Queries for reporting and analysis can run in the destination system instead of competing with the app’s operational workload.
The two additions cover different parts of production work. The prompts inspect and report on a live Supabase project. Data streaming sends information to a separate analytics platform for heavier queries.
Frequently asked questions
What are Supabase production-check prompts?
They are prompts that let an AI agent inspect a Supabase production project for health, security, performance and resource use. The agent can query logs and report its findings inside Claude, Codex or Cursor.
What controls apply when an agent connects to Supabase?
An organisation can control connections through its identity provider and restrict tokens to selected projects. The agent asks before creating a paid project or executing SQL that could delete data.
When is Supabase data streaming available?
Data streaming to BigQuery, ClickHouse, DuckLake and Snowflake is in public alpha. The source does not list plan availability or pricing.
Sources
4 checkedHow we cover tool news: Create With's tool desk drafts these reports with AI from the sources listed above and checks them against those sources before publishing.
Xano
ElevenLabs
OpenClaw
Bubble