LLM apps in prod need more than accuracy—they need cost visibility.
Tokens drive spend, but most systems lack insight.
Datalayer’s open-source agent-runtimes adds runtime-level tracking, OTEL pluggability & React charts—works with any LLM.
PR: github.com/datalayer/ag...
📄🌍 Published in #IEEE GRS Magazine (2026): “Artificial Intelligence for Earthquake Response.”
Datalayer ranked 2nd overall 🥈 in ESA + International Charter challenge.
Paper: ieeexplore.ieee.org/document/113...
DOI: doi.org/10.1109/MGRS...
We just opened a Discord for builders working with Datalayer AI Agents 🧠⚡
No marketing fluff—just real workflows, early features, and direct access to our team.
If you’re building or exploring, come join us 👇
discord.gg/YQFwvmSSuR
🚀 Jupyter MCP Server v0.23.0 is out!
Better notebook interactions, improved tool consistency, and more reliable execution ⚡
Thanks to all contributors 🙌
🔗 github.com/datalayer/ju...
#Jupyter #AI #MCP
Datalayer is writing a paper on cost-effective LLM agents (70–92% token reduction via codemode).
Seeking co-authors with real-world agent data.Read & apply → datalayer.ai/research/con...
#LLMAgents #AIEngineering #OpenScience
Our new Datalayer variant is built for AI Agent workloads—cutting data costs by up to 90% without sacrificing performance.
Scale smarter, faster, and cheaper! 💡
🔗 datalayer.ai
#AI #DataInnovation #AIAgents #CostEfficiency
⚙️ Run Python notebooks instantly in VS Code with the Datalayer Pyodide kernel — no Python install, no Conda, no setup.
Open a notebook and start experimenting ✨
🎓 Beginners | ⚡ Quick tests | 🔒 Sandbox-safe
Read more 👉 medium.com/datalayer/st...
#VSCode #Python #Jupyter #Pyodide
6/6 😴 SLEEP AT NIGHT
Production AI is scary. What if it fails?
Agent Runtimes handles it:
• Automatic retries when tools fail
• Health monitoring you can query
• Graceful degradation
⭐ github.com/datalayer/ag...
📝 datalayer.blog/2026/01/13/a...
5/6 🔄 FREEDOM TO CHOOSE
Locked into one AI provider? Not anymore.
Switch between OpenAI, Anthropic, Azure, or Bedrock without changing your code. Pick the best model for each task.
Your agents. Your choice.
4/6 💼 REAL-WORLD TASKS
Your agents can now:
• Search the web for current info (not just training data)
• Research people & companies on LinkedIn
• Read and write files
• Call any API
One config file. That's it.
3/6 🔌 INDUSTRY STANDARDS
Today's updates introduce:
• Industry-standard transport (#AGUI, #A2A, #A2UI, #VercelAI)
• Rich UI extensions (progress, forms, dashboards)
• First-class Model Context Protocol (MCP) support
Build on protocols that last.
2/6 🎯 WORKS EVERYWHERE
Our first user is the Jupyter AI Agents extension — bringing intelligent agents directly into JupyterLab notebooks.
But Agent Runtimes is a general-purpose framework. Use it in web apps, CLI tools, or anywhere you need AI that takes action.
🧵 1/6 What if your AI agent could actually DO things?
Not just chat. Actually search the web, access databases, read LinkedIn profiles, execute code.
That's what we built. Here's what you can do with Agent Runtimes 👇
Adding #OTEL to MCP Compose with the kind help of @logfire.pydantic.dev
github.com/datalayer/mc...
Less glue code. More React. More Jupyter.
#React #Jupyter #DevTools #OpenSource
This bridges the gap:
- frontend engineers → data & kernels
- data engineers → real React UIs
Jupyter UI supports the new Jupyter Embed, making notebooks and kernels first-class in web apps.
We now ship real, live React examples — full apps you can run, inspect, and reuse.
We simplified the structure so React devs can embed Jupyter without learning Jupyter internals.
1/ Jupyter UI documentation got a refresher
If this makes sense to you, you’re the target audience 👇
<JupyterReactTheme>
<Notebook path="analysis.ipynb" />
</JupyterReactTheme>
jupyter-ui.datalayer.tech
🚀 Jupyter MCP Server docs just got a refresh!
Flattened TOC, new Reference & Providers sections, added JupyterHub + Google Colab providers (WIP), and more on security & multi-user use cases.
Docs: jupyter-mcp.datalayer.tech
Feedback welcome! 💬
🪐 Jupyter Embed is here!
Add interactive Python cells, notebooks & terminals to ANY website with a simple script tag.
🎯 Live demo: jupyter-embed.datalayer.tech
Built with ❤️ for educators, developers & data scientists.
#Jupyter #Python #OpenSource
The Jupyter UI docs just got a fresh revamp ✨
Now with embedded Storybook—take a look!
👉 jupyter-ui.datalayer.tech
Jupyter React 2.0 is live! 🎉
Breaking changes: Notebook2 is now default, kernel setup moves to useJupyter(). See migration examples github.com/datalayer/ju...
1.5 days with #AI instead of maybe 1.5 weeks: Vite.js support and a revamped Jupyter React examples. AI is redefining software development timelines.
Devil’s in the details. This is why you’ll see a spinner in the VS Code runtime launcher while the remote runtime spins up.
Datalayer is joining the AgentX AgentBeats challenge! 🥊
To benchmark our data agent, we "agentified" DABench using the A2A protocol. The setup uses Pydantic AI and our Jupyter MCP Server.
We think A2A + MCP is the future of agent interoperability. 🤝
github.com/datalayer-ch...
#AgenticAI #MCP
The Datalayer VS Code extension is here. It brings a Notion-style editor (built on Lexical) right into your Jupyter environment.
Just code, write, and explain in one place.
It's open source 👐 and integrated with Copilot 🤖.
medium.com/datalayer/no...
#VSCode #Jupyter #DataScience #OpenSource
A new UI style more like Notion/Google Docs for Data Analysis, with AI Agent
