> Full Neon documentation index: https://neon.com/docs/llms.txt
# Postgres version updates, Claude Fable 5.2 and Kimi K3.1 Pro, Stripe data sync, and more
## Postgres minor version updates
We updated supported Postgres versions to [18.7](https://www.postgresql.org/docs/release/18.7/), [17.12](https://www.postgresql.org/docs/release/17.12/), [16.16](https://www.postgresql.org/docs/release/16.16/), [15.20](https://www.postgresql.org/docs/release/15.20/), and [14.25](https://www.postgresql.org/docs/release/14.25/). This security release fixes 52 vulnerabilities. See the [Postgres release announcement](https://www.postgresql.org/about/news/postgresql-187-1712-1616-1520-1425-and-19-rc-2-released-3442/) for details.
Neon [applies the update for you](https://neon.com/docs/manage/updates) the next time your [compute](https://neon.com/docs/manage/computes) restarts, so there's nothing you need to do. The one exception is fixed-size computes larger than 16 CU, which don't get scheduled updates, so [restart those yourself](https://neon.com/docs/manage/computes#restart-a-compute) to pick up the fixes. To learn how we handle minor version updates, see our [Postgres version support policy](https://neon.com/docs/postgresql/postgres-version-support).
To see which version you're on, run this in the [SQL Editor](https://neon.com/docs/get-started/query-with-neon-sql-editor) or any Postgres client:
```sql
SELECT version();
-- PostgreSQL 18.7 (a21afb5) on aarch64-unknown-linux-gnu, compiled by gcc (Ubuntu 13.3.0-6ubuntu2~24.04.1) 13.3.0, 64-bit
```
## Claude Fable 5.2 and Kimi K3.1 Pro on the AI Gateway
The [AI Gateway](https://neon.com/docs/ai-gateway/overview) lets you add AI features to your app, like summarizing notes, describing uploaded photos, or answering customer questions, with one Neon credential for models from many providers. To start building, see [Get started with the AI Gateway](https://neon.com/docs/ai-gateway/get-started). Two new models are available this week:
- **Claude Fable 5.2** (`claude-fable-5-2`): Anthropic's latest Fable model, with a 1M-token context window. Call it through the chat completions or native Anthropic Messages endpoint. See [Claude Fable 5.2](https://neon.com/docs/ai-gateway/models/claude-fable-5-2).
- **Kimi K3.1 Pro** (`kimi-k3-1-pro`): Moonshot's successor to Kimi K3. See [Kimi K3.1 Pro](https://neon.com/docs/ai-gateway/models/kimi-k3-1-pro).
Try either one with the AI SDK:
```typescript
import { neon } from '@neon/ai-sdk-provider';
import { generateText } from 'ai';
const { text } = await generateText({
model: neon('claude-fable-5-2'), // or 'kimi-k3-1-pro'
prompt: 'Explain database branching in one sentence.',
});
console.log(text);
```
See the [model catalog](https://neon.com/docs/ai-gateway/overview#available-models) for capabilities and pricing.
## Sync Stripe data into Postgres from the Integrations page
The **Integrations** page in the [Neon Console](https://console.neon.tech) now has a **Stripe** card. Select **Add** to open [Stripe Data Pipeline](https://docs.stripe.com/data/data-pipeline/real-time-sync-to-postgres), connect your Neon account, and create a pipeline. From then on, Stripe keeps your customers, subscriptions, invoices, and charges in sync in your Neon project. You don't write webhooks or sync code.
Once your billing data sits next to your app data, you can gate features on a customer's plan, show invoice history in your app, follow up on failed charges, or report on revenue with SQL. See [Sync Stripe data to Lakebase Postgres on Neon](https://neon.com/guides/stripe-data-pipeline-neon).
## Project permissions for Object Storage, Functions, and AI Gateway
[Object Storage](https://neon.com/docs/storage/overview), [Functions](https://neon.com/docs/compute/functions/overview), and the [AI Gateway](https://neon.com/docs/ai-gateway/overview) now check each member's [project role](https://neon.com/docs/manage/user-permissions) (Viewer, Editor, or Admin) before allowing an action:
- **Object Storage:** Downloading objects or creating a presigned download URL requires Editor. Viewers can still list buckets and objects. Creating a public bucket requires Admin.
- **Functions:** Registering or deleting a custom domain requires Admin.
- **Credentials:** Creating, revealing, rotating, or revoking Object Storage and AI Gateway credentials requires Editor. Viewers can see credential metadata only.
Organization API keys aren't affected, so automation keeps working. See [Per-project permissions](https://neon.com/docs/manage/user-permissions#per-project-permissions).
## New Neon Functions templates
We added three new templates to the [Function Registry](https://neon.com/functions/templates) this week. Each one is a ready-to-deploy starting point for a [function](https://neon.com/docs/compute/functions/overview):
- [`jev-triage`](https://neon.com/functions/templates/jev-triage): Route support messages by topic, urgency, churn risk, and spam with the Jev decision model on the AI Gateway.
- [`mcp-use`](https://neon.com/functions/templates/mcp-use): Build an authenticated [MCP server](https://neon.com/docs/compute/functions/mcp-server) with [mcp-use](https://github.com/mcp-use/mcp-use), with Postgres-backed event tools you can adapt (contributed by Manufact).
- [`jalco-drop`](https://neon.com/functions/templates/jalco-drop): Share files, folders, pastes, and screen recordings as links, stored in [Object Storage](https://neon.com/docs/storage/overview), with expired drops cleaned up on a schedule.
List the templates and start a function from one with the [Neon CLI](https://neon.com/docs/cli):
```bash
neon functions templates list
neon functions new jev-triage
```
Have a template to share? Open a pull request on [neondatabase/examples](https://github.com/neondatabase/examples/blob/main/registry/CONTRIBUTING.md). The contributing guide covers the template format, how to test it locally, and what we check in review.
## Update your agent skills
We updated the Neon [agent skills](https://neon.com/docs/ai/agent-skills) for this week's changes and added new and expanded skills for building live apps with [Neon Realtime](https://neon.com/docs/realtime/overview) and securing MCP servers and OAuth clients with [Managed Better Auth](https://neon.com/docs/auth/guides/plugins/oauth-provider).
To get the latest, run this in your project. It installs the skills if you don't have them and updates them if you do:
```bash
npx neon@latest skills
```
In ChatGPT, the same updates are in version 3.0.1 of the Neon plugin. See [Set up Neon in ChatGPT](https://neon.com/docs/ai/agent-setup/chatgpt).
## Fixes & improvements
**AI Gateway**
- Inkling is no longer available on the [AI Gateway](https://neon.com/docs/ai-gateway/overview) and has been removed from the [model catalog](https://neon.com/docs/ai-gateway/overview#available-models). Requests that name `inkling` now return a 400 `unknown model` error. If you use it, switch to another model in the catalog.
**Consumption API**
- Fixed an issue where the [branch metrics endpoint](https://neon.com/docs/guides/consumption-metrics#branch-metrics) left out idle and archived branches. Requests for storage metrics now return every branch billed for storage in the window, even ones with no compute usage.
**Data API**
- [Get Data API details](https://neon.com/docs/data-api/manage#get-data-api-details) now returns `auth_provider` (`neon_auth` or `external`), plus the `jwks_url`, `provider_name`, and `jwt_audience` for external providers, so you can see how a [Data API](https://neon.com/docs/data-api/get-started) verifies tokens.
**Drizzle Studio**
- We updated the Drizzle Studio integration behind the [**Tables** page](https://neon.com/docs/guides/tables), where you browse and edit your data, to version 1.8.0. This release fixes browser tabs becoming unresponsive while loading large databases, reworks the partitions view, and loads partitions on demand. For details, see the [Neon Drizzle Studio Integration Changelog](https://github.com/neondatabase/neon-drizzle-studio-changelog/blob/main/CHANGELOG.md).
**Neon Console**
- The [**Active queries**](https://neon.com/docs/introduction/monitor-active-queries) and [**Query performance**](https://neon.com/docs/introduction/monitor-query-performance) views on the [**Monitoring** page](https://neon.com/docs/introduction/monitoring-page) no longer load indefinitely when the selected compute isn't running. They now show that the compute is suspended or starting, and **Refresh** on **Active queries** is disabled until the compute is active.
**Neon MCP Server**
- The new `list_incidents` tool shows open Neon incidents and maintenance from [neonstatus.com](https://neonstatus.com), so your assistant can tell a Neon outage from a problem in your own setup. It's available on every connection, with no credentials needed.
- `run_sql` and `run_sql_transaction` now return 100 rows per statement by default, and say when a result was truncated. Set `max_rows` (up to 10,000) to see more. Failed queries now include the Postgres error code, position, detail, and hint.
- The new `query_loki_logs` tool lets your assistant search your branch's logs with LogQL and put events from requests in the same second in exact order, which helps when you debug a burst of requests or retries.
- `describe_branch` now returns the database structure for roles that can't create objects in the `public` schema.
- Fixed [Codex CLI](https://neon.com/docs/ai/agent-setup/codex) sign-in to the hosted MCP Server, which failed with "Invalid redirect URI".
See [Neon MCP Server](https://neon.com/docs/ai/neon-mcp-server).
**Realtime**
- Fixed an issue where [Realtime](https://neon.com/docs/realtime/overview) could fail to start on a branch without a `neondb_owner` role. Realtime now connects as the role that owns your database.
**Vercel**
- You can now rename a project created through the [Vercel-Managed Integration](https://neon.com/docs/guides/vercel-managed-integration) from the Neon Console or API. Neon renames the matching Vercel resource too, so the names stay in sync.