Publisher
Stories with reporting from Neon, ranked by what is gaining ground now. Each story links to the original articles.
NeonNeon makes its backend suite generally available, combining Lakebase Postgres, Object Storage, Functions, Managed Better Auth, and AI Gateway. The company says the tools are ready for production and designed as backend primitives for agents.
NeonNeon releases the beta of Lakebase Search, combining vector and full-text retrieval through the `lakebase_vector` and `lakebase_text` Postgres extensions. Its design uses IVF and RaBitQ, BM25 with top-K pushdown, and indexes on object storage that persist through scale-to-zero and branch instantly.
NeonNeon introduces AI Gateway, enabling applications to call language models from the Neon backend. The service is intended to consolidate model access rather than requiring developers to manage accounts with individual labs.
NeonNeon launches Neon Object Storage, an S3-compatible object store integrated with the Neon backend. It supports branching, allowing files to be included in isolated copies alongside database branches that take about a second to create.
NeonElectric, the team behind PGlite and the Electric sync engine, joins Databricks and its Neon team. The move brings lightweight WASM Postgres and real-time sync capabilities to Neon.
NeonNeon AI Gateway adds embedding models through the same OpenAI-compatible endpoint and credentials used for chat. Users can generate vectors, store them in Lakebase Postgres, and query them with Lakebase Search within one Neon branch.
NeonLakebase Postgres introduces autoscaling that removes the need to choose an instance size before the workload is known. The change targets database capacity management by adjusting to workload demands.
NeonStripe adds the `stripe projects build` command, which discovers and scaffolds wired project templates in one command. The launch extends Stripe Projects for teams provisioning their application stack.
NeonNeon increases the monthly data transfer included in every paid plan from 100 GB to 500 GB, effective June 1, 2026. The change is automatic and is intended to eliminate egress charges for most workloads.
NeonNeon Functions can now run on a schedule, extending them beyond jobs triggered by HTTP requests. The change addresses backend tasks such as pulling an external API into Postgres every 15 minutes without relying on an external scheduler or disabling scale to zero for a branch using pg_cron.
NeonVercel's CLI now installs a provider's agent skills when users add a Marketplace integration from the terminal. The command can provision a Neon database on a Vercel project and give a coding agent instructions for using it.
NeonNeon makes Object Storage, Functions, and AI Gateway available to all users in beta. The offerings include storage that branches with data, long-running compute beside the database, and one API with one bill for seven model providers backed by Databricks.
NeonNeon now lets users assign custom domains to Neon Functions, which previously used Neon invocation URLs. A function can now be reached through a domain such as `api.example.com`.
NeonNeon launches @neon/tools, which turns the @neon/sdk client into typed tools for agents. The package includes adapters for MCP, Mastra, and Eve, and supports expansion of the hosted Neon MCP Server.
NeonNeon introduces @neon/sdk, a type-safe TypeScript client for its OpenAPI-based Neon API. The client streamlines infrastructure provisioning by handling the polling needed while resources become ready.
NeonNeon changes the compute cache on large fixed-size Lakebase Postgres instances by putting most machine memory into Postgres shared buffers and backing the cache with huge pages. This keeps hot pages in DRAM rather than a local disk cache, serving the same working set faster and with less CPU.
NeonNeon adds a pure-TypeScript PostgreSQL client to neonctl so its psql command works even when the system psql executable is unavailable. The team says it uses a byte-exact conformance harness against PostgreSQL’s regression suite.
NeonClaimable Neon allows agents to provision a temporary Neon project without creating an account or providing payment details. It implements the anonymous registration method in auth.md, an open agent registration protocol authored by WorkOS.
NeonNeon Functions lets developers deploy Node.js 24 compute on a Neon branch in the same region as a Lakebase Postgres database, with DATABASE_URL injected automatically. The functions support long-running work, including agent streams and WebSockets or SSE connections that stay open for minutes.
NeonNeon introduces the `neon api` command to bridge gaps between human-friendly CLIs and underlying APIs used by AI agents. The command aims to provide predictable access without exposing secrets through direct raw-API calls.
