Base44
Build and edit Base44 apps from an AI assistant, and query data your app's users submit.
Query, inspect and write to PlanetScale MySQL and Postgres databases.
Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.
Paste it into any MCP client. Setup docs
The PlanetScale MCP server connects a PlanetScale account to Claude, ChatGPT and any MCP-compatible agent. Its 16 tools list organizations, databases and branches, read schema and query performance Insights, pull invoices, and run SQL — including a write tool that accepts INSERT, UPDATE, DELETE and DDL. Sign-in is OAuth against PlanetScale, or a service token.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
Not on a branch with safe migrations enabled. PlanetScale documents safe migrations as automatically rejecting DDL statements sent to that branch, whatever the client, and it recommends the setting for all production branches. On a branch without it, DDL is permitted and the only brake is a prompt asking the model to seek confirmation.
Use the insights-only endpoint, or a scoped service token. PlanetScale publishes a second server at mcp.pscale.dev/mcp/planetscale-insights-only that excludes both query tools. For service tokens it documents an exact read-query permission set built on `connect_production_read_only_branch`, and warns explicitly against granting `connect_production_branch`.
It blocks three specific things. PlanetScale documents the server rejecting `UPDATE` and `DELETE` statements that carry no `WHERE` clause, and rejecting `TRUNCATE` outright. DDL is not blocked at this layer — the docs say it prompts the model to request human confirmation, which is a client-side behaviour rather than a server-enforced refusal.
Yes, if the connection carries invoice permission. Two of the sixteen tools list invoices and break out line items per branch, and both need `organization:read_invoices` on OAuth or `read_invoices` on a service token. PlanetScale documents its billing page as available to organization administrators, so a non-admin connection is unlikely to reach it.
Whichever branch the assistant names in the call. Nothing in the tool set restricts writes to development branches, and PlanetScale's service-token docs describe granting `connect_production_branch` specifically so the write tool can reach production. The branch guardrails are safe migrations and the permissions on the credential, not the connector.
Yes, with a service token. PlanetScale documents sending the token secret as a bearer header to the same endpoint, and the token starts with `pscale_tkn_`. It warns that the token ID, and the `id:secret` pair, are both rejected. Claude.ai, Claude Desktop, Claude Managed Agents and Notion accept OAuth only.
Row-level security may be hiding them. PlanetScale runs Postgres read queries under a short-lived role holding `pg_read_all_data`, and states that role does not bypass RLS. On a zero-row or zero-count result the server checks for active policies and returns a warnings array with the code `postgres_rls_active` naming the affected relations.
Five dollars a month at the entry point. PlanetScale's pricing page on 2026-08-19 listed a single-node Postgres PS-5 cluster at $5 per month in AWS us-east-1, with the high-availability three-node PS-5 at $15 and Metal databases from $50. There is no free tier on the published pricing page.
.md suffix; the directory's URL returned 200 directly with no redirect) · retrieved 2026-08-18initialize probe — 2026-08-18, HTTP 401 with www-authenticate · retrieved 2026-08-18planetscale.com/robots.txt allows all paths and carries Content-Signal: ai-train=yes, search=yes, ai-input=yes — synthesis expressly permitted (fetched 2026-08-18) · retrieved 2026-08-18Connect PlanetScale once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.