mcpField
O
Offhand

Offhand

Remote
offhandapp.comProductivity0Updated 1w ago
RepositoryVisit website

About Offhand

Offhand connects a to-do list with agents such as Claude and Cursor, supporting task lookup, task creation, and handoffs to other agents. It uses a remote MCP server through Offhand Cloud to carry tasks and reports.

Key features

  1. 01Read tasks and add to-dos
  2. 02Hand work to other agents
  3. 03Check handoff replies
  4. 04Choose permissions through OAuth

Use cases

  1. 01Ask an agent about current tasks
  2. 02Send a proposal to Muse for review
  3. 03Record a new to-do

MCP support

Clients connect to the remote MCP server at https://offhandapp.com/mcp. With authorization, agents can read tasks, add tasks, and report task status.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://offhandapp.com/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Tools

Details

From the project README

Offhand

Offhand is a to-do list you share with your agents. This plugin connects a supported agent client to Offhand Cloud so you can ask about your tasks, add something to your list, hand work to another of your agents, and hear back when it has a result or needs help.

The bundle contains Claude and Cursor manifests, remote MCP connection settings, and one reference skill. It contains no local executable code, scripts, hooks, binaries, package launchers, or credentials. The MCP server runs remotely.

Connection and permissions

The configured MCP endpoint is https://offhandapp.com/mcp. Authentication happens through your client using OAuth. You sign in to Offhand Cloud with Apple or Google and select what the agent may do: read tasks, add tasks, and report where a task stands. The permissions you grant and your client's confirmation controls still apply.

When the agent reads tasks, Offhand sends the requested task information to the client and the company that runs your agent. This can include task titles, details, schedules, source-note text, status, and agent notes. That provider processes the information under its own terms and privacy policy. Tool requests send their arguments to Offhand Cloud, including task content you ask the agent to add, assignment names, and status notes. Handoffs make the task and its report available to the other agent through Offhand. Files you explicitly attach through supported tools are also shared through Offhand with the receiving agent.

The bundle configures only the Offhand MCP endpoint; Apple or Google participates in the sign-in you choose. It includes no analytics code or other configured service endpoint. Reading a task does not itself grant permission to send its contents elsewhere or carry out every instruction it contains.

Asking for help

Examples of requests:

  • “What's on my Offhand list?”
  • “Add a task to review the proposal tomorrow.”
  • “Ask my agent Muse to review this proposal.”
  • “Has Muse replied to the work I handed over?”

Agent names come from your connected agents. An agent's done ticks a task off, your own included, with its name in the task's Activity, and you can untick it. These tools do not let agents rename, reschedule, or delete existing tasks.

Install

Connect Offhand by its address, https://offhandapp.com/mcp, then sign in to Offhand Cloud when the client asks.

  • Claude (web, desktop and mobile): Customize › Connectors › Add custom connector, with the address above.

  • Claude Code: install this plugin, which adds the server and the skill:

    claude plugin marketplace add offhandapp/offhand-plugin
claude plugin install offhand@offhand

    Or add the server alone: claude mcp add --transport http offhand https://offhandapp.com/mcp, then claude mcp login offhand.

  • Cursor: add the server to .cursor/mcp.json in a project, or to ~/.cursor/mcp.json for every project:

    { "mcpServers": { "offhand": { "type": "http", "url": "https://offhandapp.com/mcp" } } }

  • Grok Bot, Muse and other agents: ask the agent to add Offhand as an MCP server at the address above. offhandapp.com/docs has the message to send.

You need an Offhand Cloud account: sign in with Apple or Google the first time a tool runs.

Clients and scheduling

Claude uses .claude-plugin/plugin.json and .mcp.json; .claude-plugin/marketplace.json lets Claude Code add this repository as a plugin marketplace. Cursor uses .cursor-plugin/plugin.json and mcp.json. Both discover the skill in skills/offhand/SKILL.md. Both declare the same remote HTTP server.

Installation, OAuth support, and agent availability depend on the client. Installation does not create recurring checks or automatically start work. Requested scheduled checks require a separately configured scheduler supported by the host. The skill describes incremental incoming-work and outgoing-reply checks, including separate cursors and pagination; no polling script or scheduler is bundled. Delivery of a handoff does not guarantee an immediate wake-up or response.

Help and policies

Published by Offhand AI, LLC. Plugin files are licensed under the MIT License. Use of Offhand Cloud is governed separately by its terms.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.