cv4vs Agents

Claude Code with full access to Visual Studio. A chat pane, a terminal, and an agent that can drive the IDE itself: build the solution, read the errors, step the debugger, follow references through the symbol graph, run the tests.

77 MCP tools, in process — so the agent reads the live state Visual Studio already holds, not the files on disk.

Visual Studio 2022 and 2026. Free, GPL-3.0.

What the agent can actually do


Build	build and rebuild, with the errors handed straight back
Diagnostics	errors and warnings from the live language service, including the ones an edit just introduced
Debugger	breakpoints, stepping, locals, call stack — and it offers to look when you stop on an exception
Tests	the Test Explorer's own tests: list, run, read the failures with their message and stack — and run them under the debugger
Navigation	go to definition, find references, symbol search
Editor	active document, selection, open files
Solution	projects, structure, references

The chat panel and the diff viewer are the easy part. Compiler diagnostics are not: they live in a structured model behind the Error List, not as text in a buffer. Reaching them means going through Visual Studio's own APIs — which is what these tools do, via Roslyn's per-document language services and EnvDTE rather than a C#-only path, so they keep working in C++, F# and TypeScript.

And it offers before you ask. Stop on an exception and a bar appears over that file — one press asks about the break, and the answer comes from the live call stack and locals rather than a guess at the source. Breakpoints are opt-in (you placed it, you know why you are there) and stepping never asks.

Before you start

This extension needs the Claude Code CLI installed separately — it drives it, and cannot work without it:

winget install Anthropic.ClaudeCode
# or: npm install -g @anthropic-ai/claude-code

Other platforms and methods are in Anthropic's setup guide. If it is missing, the pane says so and links you there instead of failing silently.

Then open View → cv4vs Agents → Claude. The IDE tools are wired up automatically — nothing to configure, no API key, no separate billing: it drives your own claude.exe, with whatever account you are already signed in with.

It is not a fork of the CLI. The binary is never bundled, and version differences are handled by feature detection rather than by pinning a version.

Two panes, one session

The chat pane docked in Visual Studio The CLI pane running the same session

Chat — streaming replies, thinking blocks, collapsible tool output, inline diffs, clickable file references (ClientEvents.cs:208 opens the file at that line), image attachments, and a composer with slash commands, an @ file picker and prompt history. Any reply can be read aloud.

Under the message box, a row for the turn itself: thinking, effort, model and permission mode, each a click away. Which model is answering, and what it may do without asking, are the two things that change a turn the most — they belong in sight, not three levels down a menu.

CLI — the real claude.exe in an embedded terminal, connected to the IDE over the same channel the official VS Code extension uses.

Both read the same session store, so a conversation started in one opens in the other — or in VS Code. Not one or the other: both, on the same conversation. Panes can run on different working directories at the same time.

Built for long sessions

Nothing is built, read or started until you look at it. The chat holds nothing in memory — the transcript is read from the session file on demand, newest page first, older pages as you scroll, and heavy blocks (images, sub-agent transcripts, full diffs) only when you open them. Services, the MCP server and the panes themselves start on first use, not on solution load.

A long session opens as quickly as an empty one.

Context and cost, visible

Statistics document-tab

A live gauge in the composer shows how full the context window is, and a full-window Statistics tab aggregates token usage and cost from your local session files: a navigable tree (All → Profile → Folder → Project → Days/Sessions) drives summary tiles, a GitHub-style activity heatmap and per-day/per-model charts.

A companion Usage tab shows each profile's live plan and rate-limit windows, and a Context usage tab breaks down how any historical session fills the model's context window.

File history document-tab

Before overwriting a file, Claude Code copies it — and nothing ever prunes those copies: they outlive the session and the transcript alike, quietly reaching tens of megabytes. The File history tab shows what they cost per project and per session, diffs any backup against the file as it is now, and deletes the ones you no longer want, orphans included. Your project files are never touched.

All aggregated locally. No telemetry.

Sub-agents and plugins

The sub-agent panel

When Claude fans out to sub-agents, the chat usually just sits there on a spinner. Here they get their own panel: how many are alive, what each one is doing right now, how long it has been at it, and a Stop button next to every one of them. Their tool calls stay grouped under the agent that made them.

A plugin manager covers the rest of the CLI's ecosystem — Installed, Available and Marketplaces tabs: install, enable or disable, add a marketplace, without leaving the IDE.

Leave the desk, keep the session

The Remote Control card and its toolbar indicator

/remote-control opens a running session to claude.ai/code and the Claude mobile app. Claude keeps running here, on your machine, with your solution and your IDE tools — the phone is a window onto it, not a copy.

The session link lands in the conversation with a QR code: scan it and the turn you started at your desk carries on in your hand. A toolbar indicator says it is on for as long as it is, and turns it off from the same place.

Profiles

Options → Profiles

Each pane can run against a different configuration: working directory, model, permission mode and environment. That includes any Anthropic-compatible endpoint — native Claude, GLM/z.ai, or your own host — so one pane can run on a different provider while another stays on your usual account, and your global settings are never touched. The IDE tools work the same either way.

If you came here from the Claude Code issue tracker

The things people keep asking for, and where they are:


Native diff review inside VS, not a terminal text stream	inline in the chat, click to open the file at the line
Build errors passed to Claude automatically	the build tools return them as file/line/message
A dockable chat panel, not a detached terminal	a real tool window — and a CLI pane too, if you want both
Breakpoint and debug state visible to Claude	the debugger tools: breakpoints, stepping, locals, call stack — and a bar over the file when you stop on an exception, one press from asking about it
Crash dumps, profiling, VS diagnostic tools	not yet — open an issue if you need it
Works with a Max subscription	it drives your own `claude.exe`, so whatever you are signed in with
Visual Studio 2022	supported, not just 2026

A diff opened in Visual Studio's own diff viewer

Requirements


Visual Studio	2022 or 2026 (17.0+) — Community, Professional or Enterprise
Claude Code CLI	installed separately, see above

Documentation and support

Full documentation, including MCP tools, options, sub-agents and architecture, is on GitHub.

Problems in claude.exe itself belong to the CLI's own tracker — this extension drives the CLI, it does not ship it.

Credits and legal

Artwork by filocorsa.

GPL-3.0-only — Copyright Corsinvest Srl. Made in Italy 🇮🇹

Claude and Claude Code are trademarks of Anthropic, PBC. Visual Studio is a trademark of Microsoft Corporation. This is an independent extension by Corsinvest Srl, not affiliated with or endorsed by either company; the names are used only to describe what it works with.