MCP server

Connect your AI assistant to Vi

Vi has a built-in MCP server. Add one address to Claude, Claude Code, Cursor, VS Code or Codex, approve once in your browser, and the assistant can read, create and work on Vi tasks as you.

What is the Vi MCP server?

The Vi MCP server is a remote Model Context Protocol endpoint at https://useviapp.com/mcp that lets an AI assistant work in Vi with your own account. It uses Streamable HTTP with OAuth sign-in, so there is nothing to install and no token to copy. It is included in every Vi plan, including Free, with no extra seat for the assistant; the cost of the assistant itself stays with the AI service you already use.

Connector address
https://useviapp.com/mcp

How do I add Vi to my AI assistant?

Add the connector address to your assistant, then approve the connection once in the browser with your Vi account. Pick your tool:

Claude Code

Run this once in a terminal. Then open Claude Code, type /mcp, choose vi and select Authenticate. With --scope user, Vi is available in all your projects.

claude mcp add --scope user --transport http vi https://useviapp.com/mcp

Claude (claude.ai and Claude Desktop)

Open Settings › Connectors, add a custom connector named Vi with the connector address above, click Connect and approve in the window that opens. Custom connectors depend on your Claude plan; on a team plan an owner may have to add it first.

Cursor

Add Vi to ~/.cursor/mcp.json (or .cursor/mcp.json in one project). Then open Cursor Settings › MCP and sign in to vi when Cursor asks.

{
  "mcpServers": {
    "vi": { "url": "https://useviapp.com/mcp" }
  }
}

VS Code (GitHub Copilot)

Add Vi to .vscode/mcp.json in your workspace, or run “MCP: Add Server” from the Command Palette and choose HTTP. Start the server and approve the sign-in when VS Code asks.

{
  "servers": {
    "vi": { "type": "http", "url": "https://useviapp.com/mcp" }
  }
}

Codex CLI

Add Vi to ~/.codex/config.toml, then sign in once with the second command.

[mcp_servers.vi]
url = "https://useviapp.com/mcp"
codex mcp login vi

Other MCP clients

Any client that supports remote MCP servers over Streamable HTTP with OAuth can connect with the connector address. A client that only runs local (stdio) servers cannot connect directly.

What happens when I connect?

  1. 1

    Your assistant opens a browser window with the Vi sign-in page. Sign in with your normal Vi account; your password is never shared with the assistant.

  2. 2

    Vi shows which app is asking and for which workspace. Approve only if you started the connection yourself.

  3. 3

    The assistant is connected. You see it under Settings › AI access in Vi and can disconnect it there at any time.

What can the assistant do in Vi?

Read projects, teams, members and tasks, including the description, comments, subtasks and current workflow step.

Search before it creates, then create tasks and subtasks. Likely duplicates are shown first, and it tells you which project, step and person the task went to.

Add internal comments to a task. Comments added this way are never emailed to customers.

Work on a task by its number: Vi starts it in your IDE as if you pressed Start, and the assistant reports back with a comment.

Pause a task that is in progress, after you agree in the IDE, so it can move on to another task or a subtask.

Ask you to close a task. Your IDE shows the changed files, you approve, and the same Commit & Close as the Done button runs.

What can it never do?

The assistant has your rights and never more. It cannot delete anything, move a task between workflow steps by itself, change settings, members or plans, or commit and push without your approval in the IDE. A read-only access level is available, and a workspace owner can switch AI access off for the whole workspace.

Can I say “do task 142” to my assistant?

Yes — the assistant starts task 142 in your IDE, reads its description, comments and subtasks, makes the change and adds a summary comment. When you ask it to close the task, Vi relays the request to your IDE window for that project (VS Code, Visual Studio, RAD Studio or Lazarus), where you review the files and approve. A task that a reviewer sent back is handled from the reviewer's note. Claude Code also gets two ready-made prompts: /mcp__vi__do_task and /mcp__vi__plan_subtasks.

How does the assistant know which Vi project to use?

It looks for a .vi file in your repository, the same file the Vi IDE extensions use to link a folder to a project, and asks you to confirm. Without one, it lists your projects and asks which one to use.

{"tenant":"acme","projectId":12,"projectName":"Website"}

Frequently asked questions

Is the Vi MCP server free?

Yes — the MCP server is included in every Vi plan, including Free, and the assistant does not take a seat. You pay only for the AI assistant you already use.

Do I need an API key or a token?

No — you approve the connection once in the browser with your Vi account (OAuth). Your password is never shared with the assistant, and you can remove the connection in Settings › AI access.

Which AI assistants work with Vi?

Any MCP client that supports remote Streamable HTTP servers with OAuth sign-in can connect, including Claude, Claude Code, Cursor, VS Code with GitHub Copilot and Codex.

Can the assistant close or delete my tasks?

No — it cannot delete anything or move a task by itself. It can ask you to close a task, and the close happens only after you review the files and approve in your IDE.

Does Vi see my source code through MCP?

No — the assistant edits code on your own machine and Git runs locally through the Vi IDE extension. Vi receives task data and commit metadata, never source code.

Can a workspace owner turn AI access off?

Yes — the owner can switch AI access off for the whole workspace in Settings › AI access, which stops every connected assistant immediately.

Related: IDE project management · Commit & Close · security · pricing

Try it on your own tasks

Create a free workspace, add the connector address to your assistant and ask it to list your projects.

Start free