How agent-ready is Supabase?

Supabase received an overall agentability score of 27/100 on its homepage, 47/100 on pricing, and 30/100 on documentation. These scores indicate that autonomous AI agents will face significant friction when attempting to parse information, navigate flows, and complete tasks on the platform. The product demonstrates moderate machine readability (69/100) and chunking (48/100), but falls short in areas critical for agent interaction.

The most significant gaps appear in transparency (0/100), clean handoffs (0/100), and shadow UI avoidance (15/100). Control mechanisms score 19/100, status communication reaches 30/100, and defaults sit at 33/100. These low scores mean agents cannot reliably track system state, understand decision rationale, recover from errors, or navigate multi-step workflows without human intervention.

For organizations deploying AI agents that need to interact with Supabase's interface, the current implementation will require substantial human oversight or custom integration work to bridge these gaps.

To improve agentability, Supabase should prioritize the following changes:

• Add descriptive status text to all asynchronous operations (such as 'Uploading 3 of 10 files' or 'Creating database: 45% complete') instead of displaying only spinners or progress bars without context.
• Implement explicit confirmation steps for destructive actions using distinctly-labelled buttons or dialogs, ensuring agents can identify and respect these gates before proceeding with operations like deletion or data modification.
• Provide labelled pause, cancel, or stop controls for long-running tasks with accessible names like 'Cancel import' or 'Stop migration' rather than unlabelled icons that agents cannot reliably interpret.
• Expose an activity or audit log in machine-readable format (such as JSON) alongside a visible activity feed, allowing agents to track what actions the system has taken and verify successful completion.
• Support undo or override capabilities on completed actions, particularly for destructive changes, giving agents a recovery path when operations complete in unexpected ways.
• Surface confidence levels or certainty indicators on key outputs through API response fields or UI labels that distinguish verified data from best-guess results.