AI Handoff MCP for VS Code

@handoff in the chat: see where the last AI agent stopped, rate a prompt before you send it, and continue the task.

When Claude Code, Copilot, Cursor or another agent runs out of its usage limit mid-task, ai-handoff-mcp keeps the stopping point. This extension puts it in the VS Code chat.

Ask	What you get	Model / tokens
`@handoff /resume`	The brief the next agent gets: task, where the last agent stopped, next steps, files changed after the checkpoint. A button continues in agent mode.	None
`@handoff /status`	Task, next step, last checkpoint, and what the prompt gate and loop guard did	None
`@handoff /check fix it`	How the prompt gate rates a prompt, with what is missing	None
`@handoff what is left?`	An answer from the brief, by the model you picked in the chat	Your chat model

The three commands run locally with the same code as the ai-handoff hooks and CLI, so the chat and the hooks never disagree. Nothing is sent anywhere, and session transcripts are never read.

Setup

Set up the project once, so agents record checkpoints and the hooks capture the stopping point:
```
npm install -g ai-handoff-mcp
ai-handoff-mcp init --mode global
```
Install this extension, open the project folder, and type @handoff in the chat.

The checkpoint belongs to the first folder of the workspace. Data stays in ~/.ai-handoff-mcp (or AI_HANDOFF_DATA_DIR).

Requirements

VS Code 1.100 or later with the chat (GitHub Copilot) available. /resume, /status and /check work without a model; free questions use the model selected in the chat.

AI Handoff MCP

EngHamzaAboJarad

AI Handoff MCP for VS Code

Setup

Requirements