How MCP and Glue integrate with GitHub for better team collaboration

🚀 Releasing : Nice Specs – An open-source VS Code extension for Documentation-First Architecture for Your Codebase. No servers, no third party risk, local stores for keyword and contextual search. It uses code agents you installed in VS Code and already paid for. No extra cost of hosting LLM. With Coding Agents becoming the de-facto tool for coding, it is important to control the hallucination as well as make it easier for reviewers to go through large PRs. Also, it is essential to document code components to give better context to code agents for future iteration in code. Nice Specs creates documentation from the source, keeps it in the repo, and makes it incrementally maintainable. If you’re working with large codebases, onboarding new contributors, and working with code agents — this may help. 🔗 Check It Out here https://lnkd.in/gnBEmEFt ⭐ What Nice Specs Does It introduces a new doc-focused agent (@nicespecs) that generates folder-level architecture docs, component relationships, and root-level system overviews using your actual code as the source of truth. A few highlights: Documentation Guardrails: It only responds to documentation-related prompts, keeping it focused and predictable. Smart Workspace Traversal: Understands your project structure, skips noise, and organizes the codebase into “components.” Automatic Markdown Specs: Produces clean nicespecs.*.md files inside each folder, co-located with the code they describe. Parent/Child Architecture Summaries: Builds a living component tree across the repo. Incremental Updates: Only regenerates docs for changed components, saving tokens, time, and attention. Resilience & Transparency: Resume mid-run, track progress, estimate token cost before executing. All of this is built with readability in mind — for both humans and AI agents. Open for critique, feedback and collaboration on this. This is just an initial phase for making a transition towards documentation based LLM driven software development and maintenance. Lot can be done in this. On top of this, lot of other use cases can be made in the area of code reviews and automation testing for components.

Atlassian Forge CLI just got smarter with AI-assisted error debugging. I've been testing the new Forge assistant command (still experimental), and it's changing how I troubleshoot Forge app development. What it does: When you hit errors during Forge CLI operations, the assistant sends error details to Rovo or Gemini for instant analysis. You get context-aware explanations and suggested fixes right in your terminal. The use case: Forge development involves complex APIs, manifest configurations, and deployment workflows. Errors can be cryptic. Instead of switching to Stack Overflow or docs, you get immediate help understanding what went wrong and how to fix it. How teams are using it: ➡️ Junior developers get faster onboarding without constant interruptions to senior team members ➡️ Complex DDL operation failures get decoded without manual TiDB doc searches ➡️ Deployment failures get diagnosed with specific environment context How to enable it: For Rovo: forge assistant on rovo For Gemini: forge assistant on gemini To disable: forge assistant off My experience: I recently hit a DDL operation error while setting up Forge SQL migrations. The error was vague about why my CREATE TABLE statement failed. The assistant immediately flagged that I was using AUTO_INCREMENT, which Forge SQL discourages due to hotspot issues on large datasets. It explained why AUTO_RANDOM is preferred and provided the correct syntax for my migrationRunner enqueue operation. Saved me from digging through TiDB compatibility docs and multiple failed deployments. Which AI to choose: ➡️ Use Rovo if you're already in the Atlassian ecosystem and want responses tuned to Atlassian products ➡️ Use Gemini if you prefer Google's model or need broader technical context Still experimental, but worth enabling on your next Forge project. Full docs: https://lnkd.in/eqggQG6t #Atlassian #Forge #ForgeCLI #AtlassianDeveloper #DeveloperTools #AI #Rovo #Gemini #ForgeSQL #DevOps #SoftwareDevelopment #CloudDevelopment #TechTips #DeveloperProductivity

Tired of juggling tasks between Notion and Todoist? I built an n8n template that provides a realtime, two‑way sync so your tasks stay consistent across both tools — automatically. What it does: - Keeps Notion and Todoist in sync (create, update, complete, delete). - Maps priorities, due dates (date/time), status/sections and descriptions (with Notion links). - Uses Redis locking to prevent loops and race conditions, and supports batching for Todoist Sync API limits. - Provides setup helpers (OAuth for Todoist, config generator) and a daily full-sync plus realtime diff-syncs. Why it matters: - No more manual copying or missed updates — a single source of truth for your tasks. - Preserves metadata (priority, due, status) so workflows in both tools remain meaningful. - Built for reliability: retry logic, deduplication, and safe fallback behaviours. Quick setup notes: - You need a Notion database with specific properties (Name, Status, Priority, Due, Focus, Todoist ID). - A Todoist project with matching sections (Backlog, In progress, etc.) and a Redis instance for transient locks. - Clone the workflow into your n8n, map credentials, run the setup helper and paste the generated JSON into the Globals nodes. Who should try it: product teams, freelancers, productivity enthusiasts, or any n8n user using Notion + Todoist and wanting realtime sync. Curious to see it in action? The template is available in the first comment — clone it into n8n, follow the setup notes, and ping me with feedback or questions. Template link in the comments section. #n8n #Notion #Todoist #WorkflowAutomation #Automation #APIIntegration #Webhooks #TaskManagement #Productivity #RealTimeSync #TwoWaySync #BiDirectionalSync #DataSync #NoCode #LowCode #iPaaS #SaaSIntegration #ProjectManagement #NotionAPI #TodoistAPI

From Code to Collaboration: What GitHub Just Revealed About the Next Era of Development #GitHubUniverse2025! - the event was an incredible opportunity to dive into the future of development, collaboration and AI. A few highlights : Unified workflows for agents – GitHub introduced “Agent HQ: Any agent, any way you work” — enabling orchestrating agents, tasks and context across IDEs, CLI, mobile and beyond. Custom agents built for your needs – Tailored prompts, tools, and context optimized for specific dev-tasks and teams. Code quality & observability at scale – From metrics dashboards of usage to org-wide governance of agents and AI in review. Big picture for devs – The “State of the Octoverse 2025” insights showed which languages, workflows, and ecosystems are rising. What excited me most: how teams—whether startup or enterprise—can now build, secure and scale with a more integrated AI-enabled workflow than ever before. The future is less “disjointed tools” and more “single command center” for agents, code and collaboration. Key take-aways that I’m bringing back to our work: Invest in agent orchestration early—defining the “what” and “who” of automation within the team. Make sure governance, access and metrics are part of the architecture, not afterthoughts. Embrace the shift: dev workflows will increasingly blur the lines → coding + prompt-engineering + orchestration. Keep an eye on ecosystems and languages—they’ll keep evolving faster than ever. https://lnkd.in/gu6EEQsb #developer #AI #software #GitHub #innovation #futureofwork #DevOps

💡 We’ve launched our Substack series! Beyond product updates, this publication shares Squadbase’s thoughts, insights, and behind-the-scenes perspectives on Vibe Coding, Business Intelligence, data utilization, and MCP (Model Context Protocol). We’ll post new editions every Monday — diving into topics that shape the future of AI-driven work and BI development. Here are the key highlights: - Vibe Working is rising — The concept inspired by Vibe Coding is spreading from engineering to all areas of knowledge work. Even Microsoft is embracing it through new AI agent features in Office 365. - Claude Skills reshape AI extensibility — Anthropic’s new Claude Skills introduce a lightweight, script-based way to extend AI capabilities without heavy context or complex setup. - Claude Code gets smarter — The latest updates (v2.0.20–2.0.22) balance cost and capability by using Haiku for execution and Sonnet for planning, with a redesigned “Plan Mode” UX. - Microsoft backs open-source MCP projects — GitHub Copilot(GitHub) and VS Code teams have sponsored seven MCP initiatives, signaling MCP’s move into the mainstream open developer stack. 🔗 Read the full post: https://lnkd.in/gdeqZ-AH

🚀 From Code to Clicks: Integrating Page Scripting with AL-Go CI/CD Ever wished your Business Central CI/CD pipeline could actually think like a user? Not just compile AL code… but open a Sales Invoice page, post it, and verify it works — automatically? That’s now possible. With Microsoft’s new Page Scripting Tool and the @microsoft/bc-replay package, Business Central pipelines can finally replay real UI interactions — and I’ve documented the full setup. 💡 In my latest blog, I show how I: Bootstrapped an AL-Go repo and added bc-replay Recorded a real Sales Invoice posting flow Integrated bc-replay as a post-CI workflow Watched GitHub Actions simulate a Business Central user step-by-step 👉 Read it here: https://lnkd.in/dfFqh7zE 🔗 Includes full YAML workflow, setup guide, and screenshots. If you’re building or maintaining Business Central extensions, this changes how you validate releases forever. hashtag#BusinessCentral hashtag#ALGo hashtag#DevOps hashtag#bcReplay hashtag#MicrosoftDynamics365 hashtag#Automation hashtag#CI_CD hashtag#GitHubActions hashtag#ALDevelopment

Here's my story about why Claude Code is a terrible name. This is not a developer tool. It's an AI tool for everyone in every role. What it just did for our Support metrics = mind blown 🤯 We used Claude Code to analyze all of our Support interactions across channels. The context came from Pylon for end user support, GitHub for developer support, Atlassian (JIRA) for escalated tickets across a couple of different projects. There is real magic ✨ in how easily this tool can take disparate data from these various sources and with the right prompts instantly create a coherent analysis. This work used to take hours of manual data collection and mapping to produce coherent analytics, which meant we only did it occasionally. It also required engineering muscle, which is not often available at a startup for tasks like this. Now we're living in the future and with Claude Code, GTM Ops can do it in an hour. The analysis isn't perfect but it's incredibly high leverage. It allows me to understand the specific individuals on our team who are putting in the heavy lifting in Support (shoutout Alyssa Yandura Jessica Herbert Mary Dickson), it allows me to understand which customers consume the most support resourcing and why, it allows me to measure and evaluate Support quality and SLA adherence across channels, among other insights. In case useful, here are the prompts and context we used. -- (gather context) Export all tickets created in October from Pylon, Jira Board 1, Jira Board 2 (Questions) into csv.  Run a script to get discussions from GitHub via API into csv output. Put them all in a local folder and open with Claude Code. (prompt) review the 4 files in this directory and provide analysis of activity across all methods of support (prompt) Can you analyze again for a leadership summary including assignee in pylon and issue tags. Where possible associate to a customer or account across all methods. (prompt) provide this in a markdown file to copy /paste into our company handbook and share (Notion)

🚀 Spent the past week experimenting with Claude Code’s new plugin system — and it’s a glimpse of where dev tools are heading. Plugins now bundle slash commands, AI agents, MCP servers, and workflow hooks into one neat package. Once installed, everything just flows: I could type commands like “generate tests for module X” or “review recent commit”, and Claude handled them intelligently. It felt way more consistent than setting things up manually every time. Most of my coding is solo, but I can see huge potential for teams. Shared plugins could unify workflows, automate repetitive dev tasks, and bring structure to code reviews and testing across a team. Setup-wise: ➡️ Installation was smooth. ⚙️ Configuration (especially for MCP servers and hooks) needed a bit of tweaking to connect agents to the right APIs and tools. In a corporate setup with pre-configured environments, this would honestly be chef’s kiss — unless you’re the one maintaining it. 😅 The ecosystem still feels early-stage — lots of community-built plugins, mixed documentation quality, and not yet a huge “trusted” library I’d drop straight into production. 💡 My quick take: • ✅ Worth trying if you have recurring dev workflows (scaffolding, testing, code reviews). • 👥 Excellent for teams that want a unified Claude Code setup. • ⚙️ Maybe hold off if you’re running one-off projects or work in a locked-down environment. • 🧩 Not plug-and-play yet, but the payoff is real once configured properly. You can explore plugins here 👉 claude-plugins.dev Curious — has anyone else started integrating plugins into their Claude Code setup yet? What’s been your experience so far? ⸻ #ClaudeCode #AItools #DeveloperExperience #DevTools #AIIntegration #CodingAutomation #ClaudePlugins #Anthropic #SoftwareEngineering #AIWorkflows #MCP #AIToolsForDev #DeveloperProductivity #AIinDevOps

Last month we shipped updates that automate test triage, streamline merge queue builds, and add NuGet support for .NET teams (and a whole lot more): https://lnkd.in/gNbKPRcJ TL;DR: - Test Engine workflows automatically identify problematic tests and create issues in Linear. - Buildkite now handles GitHub merge queues natively, creating and canceling builds automatically based on merge group status. - Package Registries supports NuGet for .NET teams, offers flexible storage options, and integrates more easily with third-party systems. More big updates are on the way 💚

So I had to explain "vibe coding" to my wife...🙄 If you are in the tech field and are connected to me, you probably know how this term has been on fire in 2025 with the ability to write code with some simple phrases and instructions. I use GitHub Copilot with VS Code in Agent mode to generate code or use the chat operation to ask questions all the time. You give it prompts, and it create one-to-many files or even an entire solution ready to go. I recently used it to create a simple solution to update Dynamics 365 easily so I could spend less time on mundane tasks. So what does this mean for future programmers? Here are some examples of tuning that was needed: 1️⃣ I want to give the code a list of milestones to automatically update for me. Push them out a month. Have you tried to paste a list of items into a command prompt? Each row is treated as a separate interaction so I had to tell it to use the clipboard instead. It then tried to add a bunch of Python libraries to my PowerShell script 😄 As a programmer, I could easily tell where it lacked because of the code it generated and what it was recommending. NOTE: I tried multiple AI models 2️⃣ The UI was spitting out some very odd lines. I found the line in the code and knew right away that was not valid syntax. I highlighted the row and asked Copilot to "/fix" this code because it isn't valid syntax. Now it knew right away it was wrong and quickly fixed it 3️⃣ While the code generator thought it would be cute and add a bunch of emojis to my code, it forgot about ANSI and other Unicode things that don't like those non-printable characters. I had to ask it to not include those for my solution and use words instead AI generation is amazing and can save a ton of time, but you still need the artists to paint the picture or the architect with the vision. I see this more of a companion and not a replacement. Happy vibe coding! #GitHub #Copilot #VSCode