Max – a federated data query layer for AI agents (and humans)
Hey HN! I built a thing and I'm really excited to share it. EDIT: I meant to link to the github, not the website: https://github.com/max-hq/max Like many of us here, I've been commonly reaching for a pattern of "pull data into db; give it to claude" for a while, whilst doing data spelunking or building tooling - for the same reasons mentioned by thellimist over here [1] and a few other recent "CLI vs MCP" posts. To that end, about a month ago I started building a project called `max` - its goal is to cut the middleman and schematise any data source for you. Essentially,…
What it does
In the maker’s words, at launch
Hey HN! I built a thing and I'm really excited to share it. EDIT: I meant to link to the github, not the website: https://github.com/max-hq/max Like many of us here, I've been commonly reaching for a pattern of "pull data into db; give it to claude" for a while, whilst doing data spelunking or building tooling - for the same reasons mentioned by thellimist over here [1] and a few other recent "CLI vs MCP" posts. To that end, about a month ago I started building a project called `max` - its goal is to cut the middleman and schematise any data source for you. Essentially, provide a lingua-franca for synchronising and searching data. In short: Max exposes a CLI for any given data source, and mirrors it locally. As in, puts that data right next to the agent. It means search is local and fast, and ready for cut, sed, grep, sort etc. More concretely: > max connect @max/connector-gmail --name gmail-1 > max sync gmail-1 > # show me what data i can search for > max schema @max/conector-gmail > # do a search > max search gmail-1 --filter"subject ~= Apples" --fields=subject,from,time I've built a few connectors over at `max-hq/max-connectors` - but the goal is that they're easy to create (sync is done via graph walk - max makes you provide field resolution so it can figure out how to sync). In practice - I've found that telling claude to run "max -g llm-bootstrap" to get acquainted, and then "make a connector for X" also works pretty well :). There's a lot still to come(!) - realtime, somewhere to host connectors, exposing and serving max nodes... I'll be updating the roadmap over the next couple of days - but I didn't want to wait any longer before sharing here. (on that note - max is designed for federation. The core is platform agnostic) In terms of what this approach makes possible - I ran a benchmark on a challenge (it's the one on the website) asking claude to find me names of a particular form from a fairly chunky hubspot (100k contacts). The metrics are roughly what you'd expect from putting the data local and avoiding any tokens hitting claude's context window: MCP: 18M tokens | 80m time | $180 cost Max: 238 tokens | 27s time | $0.003 cost (I'll explain how these numbers were calculated in a new reply) It's still early (alpha) but if you're building agents or just want local data, please try it and tell me what breaks. Thanks! [1] https://news.ycombinator.com/item?id=47157398
Does the same job
all alternatives →- ISI started a repo for sharing algorithm implementations2013 · github.com · ▲52
Everything that would be here is in the README. I hope this gets big, it has tons of potential.
- SOState of AI Agents 2024 – 184B tokens · 786M runs analyzed2024 · langbase.com · ▲6
hello everyone, my first post! AA here, founder of ⌘ Langbase.com — we are a developer platform for building and scaling serverless AI memory agents. I know surveys can be boring, but this one’s different—it’s interactive! That's very much intentional. My team and I have been up for the last 21 hours putting together this report. This was a looot of work, so I hope y'all like it. Introducing … State of AI Agents 2024 report On Langbase, we processed 184 billion tokens and handled 786 million AI agent runs from 36K developers. From all that data plus insights from 3.4K builders who filled out…
- DODataHub, open source datasets for Artificial Intelligence2017 · ▲14
Hi HN! I am an undergrad student trying to build interesting things with AI. Recently, I was looking for a dataset I could use for a new project. I realized that it is really frustrating to go through all the government websites (with terrible UX) just to find some usable dataset. I set out to build a GitHub for datasets, named DataHub. Right now, we have more than 1000 datasets from Montréal and New York City, with more cities coming soon (and possible government agencies). All of this is wrapped into a powerful search. It's a breeze to find a dataset to work on. I'd be interested to know…
- YAYorph AI – a data engineer in your pocketNov 2025 · yorph.ai · ▲6
Hey folks! I'm a founding engineer at Yorph AI, an agentic data platform, built using ADK, that helps users (starting with product managers and analysts) join data from different sources (upload or sync), build version-controlled and reliable data workflows, and clean, analyze, and visualize data — all in one place. We're also releasing semantic layer creation later this week. The beta is live at yorph.ai/login — would love to hear your thoughts and feedback! (FYI: We're still waiting on Google app verification — you'll see a warning for a few days. Dropbox shows a similar one since…
- IBI "built" Sana, a vector DB on object storageJun 2026 · github.com · ▲5
Disclaimer it is a heavily AI assisted project. The goal was not to be the most performative but the kind that's easier to learn from. I wanted to share this in case there are people who had the same idea or wanted to see something like this.
- WTWas tired of drowning in HN comments, so I built an AI Chief of StaffJan 2026 · hnsignals.com · ▲5
I've been lurking on HN for years. You know the drill: interesting headline, 200+ comments, you dive in thinking "I'll just skim for 5 minutes"... and an hour later you're 36 chambers deep in a thread about memory allocation patterns in Postgres and you've completely forgotten what the original article was about. I don't just want a "summary" (which usually just shortens the noise). I want the meta-consensus: "What is the actual trade-off being debated? Who is winning the argument? Why does this matter?" So I built HNSignals. Think of it less like a "summarizer" and more like a Chief of…
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, March 2026
the whole month →
Switch from ChatGPT to Claude with import memory feature
AI · Mar 2026 · claude.com
