Native cloud agent support in Cursor Agent CLI

Ideas Feature Requests
,
1

Feature request for product/service

Cursor CLI

Describe the request

The CLI already makes sending work to a Cloud Agent easy — prepend & to a message and it hands off (docs). But picking the conversation back up is forced onto the web UI or mobile. There’s no CLI path to list, resume, follow up, or stream a cloud agent (bc-…) the way
agent ls / agent resume work for local chats.

Please make cloud agents a first-class CLI surface, not just a one-way handoff.

Proposed commands (shape can vary):

agent cloud ls                          # list my cloud agents + status / PR links
agent cloud resume <bc-id>              # attach to an existing cloud agent interactively
agent cloud send <bc-id> "follow-up"    # one-shot follow-up (scriptable)
agent cloud cancel <bc-id>

And/or extend existing commands so agent ls / agent resume see cloud agents alongside local ones (filterable, e.g. --runtime cloud).

Why this matters:

  • The handoff docs already sell a CLI → cloud workflow, but the return path breaks the promise of staying in the terminal.
  • Power users and CI already use the Cloud Agents REST API; native CLI would remove the need for curl wrappers / custom SDKs for the common case.
  • Desktop, web, Slack, GitHub, Linear, and the API can all kick off / interact with cloud agents — CLI is the odd one out for the full loop.
  • Related asks (CLI↔IDE session sync) don’t cover this: those are about local session stores. This is about bc- cloud agents specifically.
    Current workaround: REST (POST /v1/agents, POST /v1/agents/{id}/runs, SSE stream) with a CURSOR_API_KEY. Works, but shouldn’t be required for what & already starts.

Operating System (if it applies)

all

5

Hey @Dylan4

Before rolling your own curl wrappers: the TypeScript SDK already covers the full loop against the same bc-… id & gives you:

import { Agent } from "@cursor/sdk";
await Agent.list({ runtime: "cloud" });
const agent = await Agent.resume("bc-...");
await agent.send("follow-up");
await Agent.cancelRun(runId);

Cookbook starters if useful: coding-agent CLI, agent kanban.

Still not the native agent cloud ls you want - fair ask, and we’ll be tracking this thread to gauge interest. If anyone has a preference between interactive attach vs scriptable one-shot, chime in.