Netlify adds 150+ integrations and persistent Node.js backends
Netlify's November 18 redesign links prompt-based coding, Git workflows and 150-plus business-data integrations. Its persistent Node.js backend targets chat and agent workloads.

Why it matters
- Netlify expects $130 million in ARR by the end of 2026, nearly 130% year-over-year growth.
- Users identifying as new to AI building increased more than 400% over six months, alongside up to 100,000 daily signups.
- In 30 days before launch, more than 500,000 users made over 1.7 million agent runs, producing over 700,000 production deploys.
- Connectors add more than 150 integrations, including Notion, Stripe, Airtable, Linear, Salesforce, GitHub and Snowflake.
- Netlify Server entered Early Access on November 10, 2026 and drew more than 10,000 new projects in its first few days.
By Rebecca Stone
Netlify expects annual recurring revenue to reach $130 million by the end of 2026, nearly 130% year-over-year growth, as it expands its AI app-building platform.
The company announced its redesigned app-building experience on November 18, 2026. The release adds Connectors for Netlify's AI Agent Runners and brings prompt-based creation into the same workflow as conventional development.
Netlify says the expansion targets people who understand a business problem but may not write code, while retaining Git-based workflows and production controls for developers. The company reports a more than 400% six-month increase in users identifying themselves as new to AI building, alongside up to 100,000 daily signups.
Netlify's existing Git and deployment model supplies the production layer for that expansion. The announcement presents agents as another way to operate the platform, not as a replacement for source control or release review.
The redesign targets two groups with different ways of working: domain experts and developers. Netlify must turn that broader usage into recurring production workloads to support its revenue forecast.
How do prompts and Git work in one project?
The new interface starts with a description of the desired application and shows agent-made changes in a live preview. A user and the agent can ask questions before Netlify implements a change, giving both sides a chance to clarify the requested behavior.
Users can inspect or restore earlier versions. Publishing remains a separate decision from implementation, so Netlify does not send every update live automatically. The separation gives prompt-driven iteration a review step before a release reaches production.
A domain expert can iterate through prompts while a developer works on the same project through Git, an integrated development environment, the command line or another coding agent. Netlify presents this as one application shared across different tools, not a separate prototype that must be rebuilt by hand.
Netlify's Agent Runners documentation lists these supported agents:
- Claude Code
- OpenAI Codex
- Google Antigravity
- Meta Muse
- OpenCode
Build mode changes code and creates a Deploy Preview for review. Ask mode answers questions without changing code and without producing a preview.
In the 30 days before the announcement, more than 500,000 users made over 1.7 million agent runs. Those runs produced more than 700,000 production deploys, according to Netlify.
What business data can agents use?
Connectors add more than 150 integrations, including Notion, Stripe, Airtable, Linear, Salesforce, GitHub and Snowflake. They give agents authorized access to business tools and data without requiring a user to copy a specification into a prompt.
In the Notion example, a user supplies a document URL and authorizes the account when prompted. The agent can then read that document and use it in the project.
Netlify also gives two concrete examples for commerce and data work: an agent can use Stripe products and prices to build a checkout flow, or query Snowflake to produce an interactive dashboard.
Each connection belongs to the individual user who authorized it. People can reuse those connections across their own projects, and the agent remains limited to the permissions granted to that account.
Joining a shared Netlify project does not expose a teammate's connected account. A developer's own credentials do not become part of another user's project merely because both people can work on the same code.
The initial release limits connected-account use to building and testing with Agent Runners. A deployed application does not automatically receive continuing access to the service; Netlify says that capability will arrive later.
Why does Netlify need a long-running backend?
Netlify introduced the underlying Netlify Server separately in Early Access on November 10, 2026. It provides a long-lived Node.js backend process rather than a short function call that ends after one request.
Netlify says Server instances can handle concurrent requests, keep persistent WebSocket connections open and reuse in-memory resources such as connection pools and caches. Those capabilities support collaborative applications, chat interfaces and agent loops that wait for model responses.
The operational difference is duration. A short, independent function invocation has a narrow lifetime, while a Server process remains available to hold state and serve longer workflows that do not fit that execution model.
That makes Server relevant to the new agent interface as well as conventional applications. An agent loop can wait for a model response and reuse a database connection without rebuilding that state for every invocation.
Netlify provisions, deploys, scales and monitors the instances. Each branch deploy and Deploy Preview can run its own backend version and database branch, which lets teams review API, schema and interface changes together.
Netlify reports that more than 10,000 new Server projects were created during the first few days of Early Access, mostly through agent runs. The figure ties the backend release to the broader agent workflow rather than presenting it as a separate developer tool.
Server usage is free during Early Access, while other metered features remain billed as usual. Servers can use up to three instances, although Free and Personal plans are limited to one instance.
An idle Server scales to zero after five minutes without requests. At general availability, Netlify plans to meter reserved memory and active CPU time, and the company says it will announce the rates beforehand.
Who can use the redesign now?
The redesigned experience is rolling out gradually. Netlify's rollout documentation says access began with most new users at launch, while other users cannot yet opt in directly.
Those users can join a waitlist, and Netlify says migration guidance will come later. Connectors for Agent Runners are rolling out to all plans, including Enterprise customers with credits enabled.
Can Netlidesk show the pattern working?
Netlify presented its internal Netlidesk help desk as an example of the intended collaboration. CISO Mark Dorsi built the first functional version in six hours, then worked with a developer to reach production in six weeks, according to the company.
The six-hour and six-week figures separate fast prototype construction from the work required to harden an internal tool for production use.
Netlify reports that the support backlog fell from approximately 300 open tickets to 70 during the first month. The company says 45 automations now handle roughly 80% of routine support work.
The example follows the sequence Netlify wants other teams to adopt: a domain expert starts an application, a developer helps make it production-ready, and the platform hosts the result. It also shows how access to company data can remain tied to the person who authorized each service.
The commercial test is whether outside teams can reproduce that pattern as access expands and deployed-app Connector support arrives. Netlify says the goal is to turn heavy agent usage into a larger base of production applications and durable recurring revenue.
Netlify's next test is whether broader access and deployed-app Connectors can turn record agent runs into durable production workloads before the company reaches its $130 million ARR target at the end of 2026.
Source: unite.ai
More from Rebecca Stone
Show full bio
Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.
754 articles
Related articles
- Entire launches agent-first Git platform with automated evals
- Microsoft merges coding and productivity AI into one Copilot app
- OpenAI's Codex Hits 4 Million Weekly Developers, Launches Enterprise Push
- Two Engineers Built Stack Overflow for Agents. ChatGPT Tripled Its Signups
- NetSuite Next is fast to deploy; CFOs are slower to adapt