Neon

Neon

Let your AI manage Neon serverless Postgres: projects, branches, compute endpoints, databases, roles, schemas, operations and consumption. It can create branches and start or suspend computes after your confirmation. Never returns passwords. With a Neon API key.

7-day free trial · No credit card · EU-hosted

Ask Neon anything

Try a real prompt against Neon via AnythingMCP — Claude calls the tools, returns the answer. Press Enter on your own message to install on Cloud.

Install Neon on Cloud
Neon · live via MCP
Share
Opus 4.7

Claude is AI and can make mistakes. Please double-check responses.

Setup in 2 minutes

  1. 1Start a free trial on cloud.anythingmcp.com.
  2. 2Click Neon in the connector store. Paste credentials when prompted.
  3. 3Mint an MCP API key and point your AI client at the generated MCP URL.
Install Neon now

Guides for Neon

Step-by-step setup for the most popular AI clients.

Setting up Neon

From the connector definition itself: where to find the credentials, what they need to be allowed to do, and which API this adapter talks to.

Neon

This connector manages Neon serverless Postgres through the Neon API v2 (https://console.neon.tech/api/v2) with an API key: organizations, projects, branches, compute endpoints (start and suspend), databases, roles, database schemas, operations and consumption. It does not run SQL (see below).

Setup (about 2 minutes)

  1. In the Neon Console open Account settings → API keys (personal key: acts as you, in every organization you belong to) or Organization settings → API keys (organization key: only that organization's projects; a project-scoped key: only one project).
  2. Create the key, copy it at once (Neon shows it only once) and paste it into NEON_API_KEY.
  3. Install and run neon_list_organizations.

Write tools need confirmation. Before neon_create_branch, neon_start_endpoint or neon_suspend_endpoint, tell the user what will happen and wait for a clear yes.

  • A running compute uses compute hours (CU-hours), which count against the plan's allowance or are billed on paid plans; neon_start_endpoint starts that meter. Computes scale to zero by themselves after the suspend timeout.
  • neon_suspend_endpoint closes every open connection to that compute at once.
  • neon_create_branch creates a copy-on-write branch (instant, no data copied). Branches beyond the plan's allowance and their storage changes can be billed; adding an endpoint also starts a compute. Set expires_at for temporary branches.
  • Switched off at install: neon_delete_branch (deletes the branch and its computes; the root and default branch cannot be deleted).

Finding things

  • neon_list_organizations gives the org_ids. With a personal API key, neon_list_projects needs org_id for organization projects; with an organization key it is inferred.
  • neon_list_projects → neon_get_project (settings and this period's usage) → neon_list_branches / neon_get_branch (the default branch has default: true) → neon_list_databases / neon_list_roles / neon_get_database_schema. Ids: projects like shiny-wind-028834, branches br-..., computes ep-....
  • neon_list_endpoints (or neon_get_endpoint for one) shows each compute's state (active, idle), size range in CU and the host name.
  • Every change runs as asynchronous operations: the answer lists them, and neon_get_operation shows when they finish (finished, failed).
  • Lists page with cursor: pass pagination.next (branches) or pagination.cursor (projects, operations) from the previous answer.

Usage and cost: neon_get_consumption returns compute unit seconds, storage, branch counts and network transfer per project and period (hourly for the last 7 days, daily for 60 days, monthly for a year). It needs org_id and is available on paid plans only (Launch, Scale, Business, Enterprise). It does not wake computes. Project and branch details also show compute_time_seconds, active_time_seconds, written_data_bytes and data_transfer_bytes for the current period.

Passwords and connection strings are never returned. Role passwords, connection URIs and the reveal-password call are left out on purpose, so no database password reaches the AI.

Running SQL is not part of this connector. Neon's SQL over HTTP endpoint (https://<endpoint host>/sql) authenticates with a full Postgres connection string including the role's password, not with the API key, so it would put a database password in front of the AI. To query data, use the PostgreSQL connector with a dedicated read-only Postgres role on the Neon host (<endpoint host>, port 5432, SSL required). neon_get_database_schema still shows the tables and columns without a password.

Common errors

  • 401 supplied credentials do not pass authentication: wrong or revoked API key. Neon answers 401 on every URL for a bad key.
  • 403: the key may not access that project or organization (a project-scoped or organization key used elsewhere), or the plan does not include the feature (consumption history).
  • 404: wrong project, branch or endpoint id. 423 Locked: another operation is running on the project; wait for it (neon_list_operations) and retry.
  • 429: too many requests; wait a little and retry.

Out of scope here: SQL, creating or deleting projects, databases and roles, resetting role passwords, Neon Auth, the Data API, snapshots and organization administration.

Your AI is two clicks away from Neon.

Install the connector, paste your credentials, ask your AI a question. 7-day free trial, no credit card.