Articles

The article introduces Model Context Protocol (MCP), an open standard designed to solve the problem of Large Language Models (LLMs) hallucinating due to a lack of real-time, external context. Similar to Language Server Protocol (LSP), MCP establishes a client-server architecture where LLM applications (hosts) connect to MCP servers to access structured data from external tools and data sources. GitHub has open-sourced its own MCP server, acting as a source-of-truth interface between GitHub and any LLM. This allows natural language requests to be translated into structured API calls, fetching real-time data and enabling diverse applications like automating issue creation, compiling team digests, and building conversational project assistants. The architecture promotes modularity, testability, and interchangeability of components. The article provides clear steps on how to install and use the GitHub MCP server in VS Code, demonstrating its immediate practical utility for developers to build smarter and safer AI tools.

The article provides an in-depth look at the most impactful features and changes introduced in Git 2.51. It begins by detailing the new 'cruft-free multi-pack indexes' (MIDXs), explaining how Git now intelligently stores object copies to ensure non-cruft packs are self-contained, resulting in smaller MIDXs, faster write operations, and improved repository read performance, citing GitHub's own 38% MIDX size reduction. Next, it introduces 'smaller packs with path walk,' a novel repacking method that optimizes delta compression by grouping objects based on their file paths, leading to significantly smaller packfiles. A major workflow improvement is the new 'stash interchange format,' which transforms stash entries into a sequence of commits, enabling seamless export and import of stashes across different machines. Beyond these core features, Git 2.51 also brings enhancements to `git cat-file` for better submodule information, extended Bloom filter support for `git log` with multiple pathspecs, the graduation of `git switch` and `git restore` from experimental status, and the deprecation of `git whatchanged`. The article concludes with a preview of Git 3.0's future defaults (reftable and SHA-256) and updates on Git's internal development processes, such as C99 feature adoption and revised patch submission guidelines.

This article details the journey of the ITU Telecommunication Development Bureau (BDT), a UN organization, in open-sourcing its internal software through a six-month GitHub Skills-Based Volunteering project. Facing limited budgets and small teams, BDT sought to leverage the open-source ecosystem to scale its impact. The author, a GitHub volunteer, outlines a four-step process: (1) thorough research of existing open-source repositories to understand best practices, (2) refining an "open and public" mindset by sanitizing sensitive code and preparing essential documentation like "Getting Started" guides and `CONTRIBUTING.md` files, (3) carefully selecting an appropriate open-source license (BDT chose BSD-2 for its permissiveness and clear attribution), and (4) actively engaging with the open-source community by identifying "good first issues" to attract new contributors. The initiative successfully equipped BDT with the expertise to manage open-source projects, ensuring code quality, security, and community collaboration, ultimately enabling wider access to their tools and fostering a broader developer community.

This article is GitHub's availability report for July 2025, detailing a single significant incident that affected the GitHub Enterprise Importer (GEI). The incident occurred on July 28, 2025, lasting approximately five and a half hours, during which GEI migrations were in a degraded state. The root cause was identified as a GEI infrastructure component being improperly taken out of service during routine internal improvements, which subsequently could not be restored to its previous configuration, necessitating the provisioning of new resources. Following the resolution, GitHub implemented improvements in infrastructure recovery, unit testing, and validation with test data. A critical outcome for users is the requirement to update their IP allow lists to include new GEI IP ranges (20.99.172.64/28, 135.234.59.224/28) and remove old ones (40.71.233.224/28, 20.125.12.8/29) across various platforms like github.com organizations, Azure Blob Storage, Amazon S3, or Azure DevOps. Affected users have received email notifications, and customer support is available for migration-related issues.