News

From structured AI brainstorming to milestone management, pixel-level GUI verification, PDF export, interactive spec design, and ISO 29119-aligned test specification generation. Every stage of development now runs within a managed quality workflow.

Tracking alone is not enough. Approval workflows, requirement tree decomposition, verification campaigns, automated report generation, and multi-agent requirement drafting bring V-model traceability to production readiness.

The V-model traceability that handoff-mcp introduced is now visual. A map, a matrix, and an interactive mindmap let you see requirement-to-test coverage at a glance inside VSCode.

Track the full chain from requirements through design, implementation, testing, and verification — all on plain Markdown documents. This release introduces layer documents, coverage reports, change-impact analysis, and automatic suspect detection with over 20 new tools.

handoff-mcp v0.34–0.35 lets multiple AI agents work in parallel across git worktrees. This release brings that picture into VSCode: who owns what, how far along each worktree is, and whether merging is safe.

Place an AI agent in each git worktree and share task state, memory, and documents through a single storage root. Lease-based locking prevents collisions, and an overview tool gives you a bird's-eye view of the whole project.

A 607 KB embedding model is now bundled in the binary, letting memory search cross language barriers. Knowledge saved in Japanese surfaces on English queries, and vice versa. The same technique powers near-duplicate detection.

Automated AI review loops can spiral endlessly if left unchecked. These releases introduce convergence controls: all findings up front, scoped rework, reviewer micro-fixes, and automatic document reconciliation.

Documents stored in .handoff/ are now first-class citizens in the extension: a sidebar browser, a Markdown viewer, and a relationship graph that includes tasks and verification progress.

Claude Code or Codex CLI — same tasks, same plans, same memory. Delegate parts of your development loop to Codex without the project context resetting.

npm install no longer needs a Rust toolchain. Prebuilt binaries now ship for six targets across Linux, macOS, and Windows on both x64 and arm64, fixing an install that had been silently breaking under npm v12.

Full document lifecycle management, improved BM25 search for Japanese text, streamlined setup, and official Codex CLI support — six versions of updates in one post.

The estimate requirement that bulk import and bulk update used to slip past is now enforced. session-execute gains express / standard / full profiles, and the bug that counted a crash as a pass is fixed. The marketplace plugin now ships its skills.

The extension now scans a workspace recursively for .handoff directories and groups tasks by project. One project selector drives every panel, and Gantt rows can be hidden individually.

A new research-loop workflow runs technical investigation across parallel agents. Nested .handoff projects in a monorepo are now discovered recursively. The VSCode hang is fixed.

A 2 KB trained model now splits Japanese at word boundaries. There is still no dictionary. Dropping the character-bigram approach improved search and near-duplicate detection alike.

Install with /plugin install — no manual MCP config needed. Multi-session support adds fork, merge, and switch. Task loop v2 improves review efficiency.

Session fork, merge, and switch support. Inline action buttons on tasks and sessions. SVG fork graph for visualizing session branching.

All 37 tools now callable as shell commands. Memory best practices packaged as a built-in skill. Hyphenated task ID resolution fixed.

The MCP server and VSCode extension now coordinate timers. When the extension is running, MCP delegates; when absent, MCP falls back to its own timer. An authority protocol guarantees zero double-counting.

Project memory gives AI coding agents durable, cross-session knowledge — lessons, rules, and conventions — with automatic injection into every prompt. A single setup command wires it all in.

We published lexsim, a Rust crate that scores text similarity across languages without a morphological dictionary — powering memory retrieval and dedup in handoff-mcp.

Under the concept of "a record of what we've built," we reorganized the site and added a news section.

We released Handoff, an MCP server and VSCode extension that gives AI coding agents memory that carries across sessions.

We launched an online environment where you can try AELM right away in your browser, with nothing to install.

We released AELM, a VSCode extension that lets you write circuits as text and renders them as schematics automatically.

The sequel to "Why Git? I" — stepping into branch workflows, team collaboration, and everyday Git practice.

A beginner-friendly book that explains, in seven chapters, why Git matters in the age of AI.