AI Tools
Connect your AI coding agent to Supabase.
Supabase provides everything you need to connect an AI coding agent to your project: a live connection to your database and platform (MCP), portable instructions your agent can reuse (Agent Skills), a one-step bundle of both (Plugin), and copy-paste prompts for tools that don't support any of the above.
See how these tools perform on real Supabase tasks in Supabase Evals, our open-source benchmark for AI coding agents.
Pick your agent
Antigravity
Plugin + MCPExperience liftoff with the next-gen agent platform.Claude Code
Plugin + MCPWork with Claude directly in your codebase, from your terminal, IDE, and more.Codex
Plugin + MCPA lightweight coding agent that runs in your terminal.Cursor
Plugin + MCPYour coding agent for building ambitious software.Devin Desktop
MCPThe first agentic IDE. Tomorrow's editor, today.Factory
MCPA self-improving system for your SDLC.fx
MCPConnect using the Supabase MCP server or plugin.Gemini CLI
Plugin + MCPBuild, debug & deploy with AI.GitHub Copilot
Plugin + MCPYour AI accelerator for every workflow, from the editor to the enterprise.Goose
MCPYour native open source AI agent — desktop app, CLI, and API.Grok
Plugin + MCPConnect using the Supabase MCP server or plugin.Kimi Code
Plugin + MCPEngineered to drop into any dev workflow and get programming tasks done fast.Kiro
MCPMove beyond AI coding to agentic engineering.omp
Plugin + MCPA coding agent with the IDE wired in.OpenCode
MCPThe open source AI coding agent.VS Code
Plugin + MCPThe open source AI code editor — your home for multi-agent development.Warp
MCPConnect using the Supabase MCP server or plugin.
Key concepts
- MCP (Model Context Protocol): a live connection between your agent and your actual Supabase project. Once connected, your agent can call tools to query data, run migrations, deploy Edge Functions, and more.
- Agent Skills: portable, on-demand instructions your agent loads when it needs Supabase- or Postgres-specific procedural knowledge. Skills don't require a live connection, and work across different agents.
- Plugin: a single install that bundles the MCP server and Agent Skills together for a specific agent.
- Prompts: static prompt files you copy into your project for agents that don't support MCP, plugins, or skills natively.
Local projects for parallel agents
Agents that work in separate git worktrees or branches each need their own database. Otherwise migrations and seed data from one task affect another. With the experimental supabase stack commands turned on, supabase start gives each worktree or branch its own local Supabase project with its own ports and data. Agents that share one checkout and branch share a local project, so give each agent its own worktree. The commands also run without Docker on Linux and on macOS on Apple silicon, which covers agent sandboxes that have no container engine. See Running multiple local projects and Docker and native runtimes.
Building AI into your app?
The tools above are for your development workflow. If you're building AI capabilities into your own product:
Report documentation problems
When an AI agent finds that a Supabase documentation page is wrong, outdated, or missing information, it can report the problem to Supabase. Supabase reviews reports to find and fix pages.
Approve or request a report
Your agent might ask you to approve a report. If the report is useful, say yes. The agent sends the report with the Supabase MCP server or the Supabase CLI, and shows you a delete token.
To ask for a report yourself, tell your agent, for example: "Report the problem you just found."
Turn off reports
The MCP server and the CLI don't ask for your approval themselves. To stop reports:
- MCP server: Set the server's
featuresparameter to the groups you use and leave outfeedback, or turn on read-only mode. Read-only mode also hides the tool that deletes reports. For more information, see MCP server configuration options. - CLI: Tell your agent not to send reports. The MCP server settings don't stop an agent that has a shell.
Delete a report
Ask your agent to delete the report with its delete token. You can also run supabase feedback delete <delete_token> in the project directory where the agent sent the report, or add --project-ref <ref>. If the CLI was signed in when the agent sent the report, use the same account.
What a report records
Besides the agent's text, each report records the following:
- The tool and its version
- From the MCP server, the MCP client name and version
- From the CLI, whether an agent ran it, the agent name, your operating system, and your CPU architecture
- When available, your Supabase user ID and project ref