Datalayer

We’ve only just started experimenting with Agent Codemode, and the results are already uncomfortable for the “LLM and MCP are inefficient” crowd. Even with a deliberately trivial prompt — “Generate 2000 words of random text and write it to a file” — the gains in token usage and execution speed are clear. This isn’t where Codemode shines the most. It’s where it barely tries. Now imagine what happens when we apply this to real data analysis and agentic workflows, where code-first execution actually makes sense. Numbers don’t care about opinions. See the difference for yourself and join the fun on the Agent Codemode repo: 👉 https://lnkd.in/eBHce4_q #AI #MCP #Codemode #LLMContext

For a long time, running a Jupyter notebook meant one thing first: ⚙️ installing Python. Choosing a version, setting up an environment, fixing dependency issues… all before writing a single line of code. We’ve been exploring a different approach. With the new Datalayer VS Code extension, you can run Python notebooks instantly using a Pyodide-powered browser kernel ❌ no local Python ❌ no Conda ❌ no setup Just open a notebook and start experimenting ✨ This is especially useful for: 🎓 beginners getting started with Python ⚡ quick experiments and tutorials 🔒 safely running code in a sandboxed environment We wrote a short post explaining how it works, where it shines, and its current limitations (yes, there are some but and we’re working on making things easier). 👉 Read the article: https://lnkd.in/eprHiqkM #VSCode #Python #Jupyter #DataScience #DeveloperTools #WebAssembly #Pyodide #OpenSource #Productivity

🤖 Building AI agents is hard. Connecting them to real-world tools? Even harder. That's why we built Agent Runtimes — a general-purpose framework for building production-ready AI agents that do real work. Our first user? The **Jupyter AI Agents extension**, bringing intelligent agents directly into JupyterLab notebooks. But Agent Runtimes works anywhere. Today's updates introduce: ✅ **Industry-standard transport protocols** — #AGUI, #A2A, #A2UI, #VercelAI. ✅ **Rich UI extensions** — Go beyond text chat. Build progress indicators, forms, and interactive dashboards. ✅ **First-class Model Context Protocol (MCP) support** — Connect to any MCP-compatible tool or service, eg **Search the web in real-time** — Your agents use Tavily to find up-to-date information, not just training data, **Access LinkedIn profiles** — Research candidates, companies, or job postings directly through your AI assistant. ✅ **Switch AI providers instantly** — Start with OpenAI, move to Anthropic, try Bedrock. No code changes needed. ✅ **Deploy with confidence** — Automatic retries, health monitoring, graceful degradation. Production-ready from day one. Stop wrestling with integrations. Start building what matters. 🔗 https://lnkd.in/e5niAj_s ⭐ https://lnkd.in/eMHs_bUn #AI #AIAgents #Productivity #Automation #OpenSource

🪐 Jupyter MCP Tools 0.1.5 is released — powered by community requirements We’re excited to announce the release of Jupyter MCP Tools v0.1.5, bringing two highly requested features shaped directly by community discussions and feedback. 🔢 New Index Indicator The new Index indicator makes it easier to reference and reason about notebook cells programmatically. This improves clarity when navigating, inspecting, or automating workflows across multiple cells — a small addition with a big impact on developer experience. 🎯 New get_selected_cell Command With get_selected_cell, tools and agents can now precisely identify the currently selected cell in Jupyter. This unlocks more context-aware interactions, enabling smarter automations and tighter integrations with MCP-based workflows. 🤝 Built with the Community This release is a direct result of open discussions and collaboration across the Jupyter MCP ecosystem. Special thanks to everyone who contributed ideas, feedback, and code through these conversations: Community request and design discussion: https://lnkd.in/dK83DrtR Related server-side discussion: https://lnkd.in/deA4Ad7y Implementation via pull request: https://lnkd.in/dR6BpAEq Open-source moves forward when users become contributors — and this release is a great example of that in action. If you’re building AI agents, developer tools, or automations around Jupyter, give v0.1.5 a try and let us know what you’d like to see next 👇 #Jupyter #OpenSource #MCP #DeveloperTools #CommunityDriven #AIEngineering

🪄🤖 VB Transform by VentureBeat is all about orchestrating AI autonomy. I won’t be able to make it, but it’s another strong signal that 2026 is shaping up to be the year of agent orchestration. https://lnkd.in/egBjTbA7 #AI #Agents #Orchestration

✨ Introducing MCP Compose — a powerful new way to orchestrate your MCP servers! We’re excited to announce MCP Compose, an open-source project from Datalayer that brings Docker-Compose-style orchestration to Model Context Protocol (MCP) servers. It’s built for teams who are moving beyond a single MCP server and need structure, visibility, and control. 💡 What can you use MCP Compose for? 🔹 Local AI development environments - Spin up multiple MCP servers (tools, data sources, agents) on your laptop with one command, inspect them live, and iterate faster. 🔹 Agent tool ecosystems - Compose and expose tools from multiple MCP servers into a single, unified interface for AI agents — with clear conflict resolution strategies. 🔹 Protocol bridging - Run legacy or CLI-based MCP servers over STDIO while exposing them to modern clients via SSE, without rewriting anything. 🔹 Team & platform workflows - Standardize how MCP servers are started, monitored, and secured across teams using Docker, tokens, and a shared control plane. 🔹 Observability & debugging - Track logs, metrics, and server health in real time through a Web UI or REST API — ideal for diagnosing tool behavior during agent runs. 🔹 Production-ready orchestration - Deploy multiple MCP servers with authentication, monitoring, and lifecycle management — without building custom glue code. ✨ Key capabilities that enable these use cases: - Unified multi-server start / stop / monitor - REST API + modern React-based Web UI - Tool discovery and intelligent composition - Programmatic control via Python API - Real-time metrics, logs, and monitoring 📦 Get started in minutes: pip install mcp-compose mcp-compose serve open http://localhost:9456 …or dive deeper with advanced composition via Python. 🔗 Explore the project & contribute: https://lnkd.in/eFCfD5Wg #opensource #MCP #AI #DevTools #Python #LLMOps #AgenticAI #microservices

🔮 2025: The Year of the AI Agent 🤖 — 2026: The Year of AI Agent Orchestration 🪄 2025 was undeniably defined by the meteoric rise of AI agents — autonomous, task-capable models that act on behalf of users and businesses, transforming workflows and delivering real value across domains. We saw agents built for customer service, workflow automation, data synthesis, decision support, and more. They weren’t just cool demos — they started to become real products with measurable impact. But as the ecosystem scales, we’re bumping into a profound limitation: agents operating in isolation don’t scale together. Today, the promise of AI agents is constrained by a lack of standards, protocols, and orchestration frameworks: ⚠️ Agents can’t reliably share context, coordinate goals, or chain tasks across environments without brittle point-to-point integrations. ⚠️ Each new agent adds complexity, not clarity. ⚠️ Without orchestration, “agents everywhere” quickly becomes **fragmentation everywhere.” This is where 2026’s pivotal shift begins — from individual AI agents to cohesive, orchestrated multi-agent systems that can collaborate, delegate, negotiate, and deliver on complex, cross-domain goals. Agentic orchestration doesn’t just multiply capabilities… it amplifies them. 👇 Excited to see this movement live and in person? Two San Francisco-based hackathons this January are tackling this orchestration challenge head-on: 🔹 Agentic Orchestration and Collaboration Hackathon — Hosted by MongoDB, this one-day sprint on January 10 focuses on enabling agents to communicate, share context, and work together toward complex goals — building tools and protocols that will be foundational for the next wave of AI systems. Finalists will demo live at MongoDB.local SF and compete for $30K+ in prizes. Cerebral Valley 👉 https://lnkd.in/eQF9AQ3D 🔹 Agentic Orchestration Hack — Hosted by Creators Corner on January 16–17, this event brings together builders to craft end-to-end agentic systems, backed by AWS, Anthropic, Anyscale, and others, with $47K+ in prizes on the line. Luma 👉 https://luma.com/q2r9lcq6 ✨ If you’re passionate about shaping the protocols and frameworks that will define how AI agents collaborate at scale — these events are where the future is being built. Let’s go beyond isolated agents. Orchestrate intelligence. Build systems that work together, not apart. #AI #AgenticAI #Orchestration #Hackathon #Innovation #FutureOfWork

🪐 ⚛️ Jupyter UI documentation got a refresher If you can write React, you can embed Jupyter. <JupyterReactTheme> <Notebook path="analysis.ipynb" /> </JupyterReactTheme> That’s the mental model — and we just made it much easier to understand and use. What’s new: ⚙️ Simplified structure → clearer React entry points, fewer moving parts 🧪 Real, live examples → running apps, not abstract snippets 🔌 New Jupyter Embed support → kernels, notebooks, outputs inside your React app This is for: - Frontend engineers who need real data, kernels, and notebooks - Data engineers who want a real UI without rebuilding everything Docs & live React examples 👉 https://lnkd.in/eG_QPXP2 Source 👉 https://lnkd.in/eTQdT5TM Think of it as Jupyter primitives for React apps. #React #Jupyter #Frontend #DeveloperExperience #OpenSource #DataEngineering

🚀 Just revamped the Jupyter MCP Server documentation! With so many ways to vibe analyse with MCP, clarity is key. We’ve: - Flattened the table of contents (no more level-3 headings) - Added Reference and Providers sections - Introduced the JupyterHub provider and a Google Colab provider (both work in progress — exciting times ahead!) - Added more information around Security and multi-user use cases 👉 The Clients section (Claude Desktop, VS Code, Cursor, Cline, and Windsurf) remains unchanged. Check it out here: https://lnkd.in/esHGNYcS 💡 Your thoughts? Open a GitHub issue: https://lnkd.in/eJx6E3RE 💬 Or join the discussion on Datalayer Discord: https://lnkd.in/eGCXGnic Let’s make the docs vibe even better ✨