Statewright restricts which tools an AI coding agent can access per phase — read-only during planning, edit tools during implementation, limited Bash during testing. Two local 13–20B models went from 2/10 to 10/10 on a 5-task SWE-bench subset with constraints enabled.
Claude Code cannot start your .NET app. It runs dotnet build and dotnet test. That's the full extent of its runtime interaction. Live endpoints, integration tests against a running server — that's still on you.
I tested it on a Clean Architecture solution with MediatR and FluentValidation. /init mapped all layers correctly. It generated xUnit tests with WebApplicationFactory. But the moment I needed a live route tested, I was back to Postman.
Dune is a 3-button Mac keypad that remaps itself based on whichever app is in focus. VS Code shortcuts when coding, mute/camera in Zoom. You describe your workflow to Claude — Claude programs the hardware. $119 early bird ($149 retail). #MacOS#ProductivityTools
Forge wraps a local 8B model with guardrails — retry nudges, rescue parsing, response validation — and lifts agentic task completion from 53% to 99%. Same model, different architecture layer, now outperforms unguarded frontier APIs on multi-step workflows.
Poke became the first AI agent approved on Apple's Messages for Business platform on June 4. Running inside iMessage, SMS, WhatsApp, and Telegram with no separate app, it handled 100 million messages in its first three months.
Asmi AI makes actual phone calls on your behalf — navigating hold queues, IVR menus, and receptionists to book appointments or resolve issues. It calls you each morning, takes your list, and texts you when the calls are done.
164% more stories. 98% more PRs. DORA metrics: flat. Code review time: up 91%. Claude Code data from 22K+ developers tells a specific story about what's actually happening in .NET teams. 1/
The teams winning with this restructured review alongside adoption — automated quality gates, pair review on AI-generated code, explicit ownership of what stays human. 4/
Canopy runs parallel Claude Code sessions in isolated git worktrees on macOS, so two agents can work the same repo without stepping on each other's files. One-click Merge & Finish, per-session token tracking, optional container sandboxing. Free, AGPL-3.0.
Grok Build is xAI's terminal CLI that runs up to 8 parallel AI coding sub-agents, each in its own Git worktree, with plan-first execution and MCP support.
Agent Dashboard shipped June 15, 2026. One screen groups all active sessions by state — waiting, active, idle — so you can see the full picture without bouncing between terminals.
Claude Code doesn't know your .NET conventions. dotnet-claude-kit fixes that — 47 skills, 10 agents, Roslyn tooling, all pre-wired for .NET 10 and C# 14. Teams report 40–50% fewer review corrections. https://t.co/heAGGIrGmz #dotnet#ClaudeCode