In 2019 we launched Netlify Functions. In 2023, Functions 2.0 rebuilt the whole experience around Developer Experience.
Today we're taking it to the next frontier: Agent Experience.
Here's what shipped. ๐งต
Building with AI in London? Come ship with us! ๐
Build with Netlify: London โ a live workshop + hack night at @RamenClubHQ's Ramen Space in Dalston.
๐ Wed 29 July, 18:00โ21:00 BST
Workshop by @AshAmplifies + Netlify credits, drinks, and 10k credits for the winning build.
Got an idea?
Skip the slides. Skip the PRD.
@thisiskp_ built a deploy preview for #HotAppSummer with Agent Runners, handed it to an engineer for review, and kept moving.
Hear how Netlify is changing the path from idea โ working software. ๐
๐ฅ Winner 3: Literal Agent Runners by Ash Wesley
What actually happens behind the scenes when you click "Build now" on an Agent Runners project? Now you know!
Congrats to this week's winners! ๐
Want to see your build here next Friday? Ship something, submit it, and tag @Netlify.
Browse everything shipped so far:
We just shipped a new pattern for background functions. The old way, appending "background" to the filename, still works. The new way is setting background to true in your config.
Both are valid. But we want agents reaching for the new one.
So before shipping, we tested which
The scenario is a real prompt plus a grading rubric, run twice: once without our skill loaded, once with. Same prompt both times, across Codex, Claude Code, and Gemini.
Without the skill, agents reach for the classic pattern. That passes. With it, they pick up the new one.
380 apps built during #HotAppSummer! ๐ฅโ๏ธ
We launched it as a fun way for the community to build together at the start of the summer. Honest expectations were maybe 5 or 10 submissions. Then the public leaderboard turned it into an arcade high score chase.
@taniacryptid and
New models, zero setup. OpenAI's GPT-5.6 Sol, Luna, and Terra now work out of the box in Netlify Functions and Agent Runners.
const openai = new OpenAI();
model: 'gpt-5.6-sol'
That's the whole integration. AI Gateway handles auth, caching, and rate limiting for you.
Luna and Terra work the same way. Swap in 'gpt-5.6-luna' or 'gpt-5.6-terra' and deploy. All three run across every Function type and in Agent Runners. Full details and code examples in the changelog:
Testing an API used to be simple. The API works, the docs match, you ship.
Then agents showed up. You don't know which one your user is running, what context it has, or what it decides to do by default.
@seancdavis29 walks through what that breaks, and how AXIS measures it.
AXIS runs real agents against your service across different models and contexts, then scores the experience. Change the API, the docs, a skill, or the MCP server, and you can see whether agents handle it better or worse.
Free and open source:
I just built my 9th app of #HotARSummer using @Netlify Agent Runners!
๐ธ Meet Spell-a-Bloom, a custom spelling practice tool I created for my kids, with just a sentence like chatting with a friend, to help them master their words.
So happy to watch my 10yo daughter is using
