Blog
Guides

How to approve Claude Cowork tasks from your phone

Answer Cowork's questions and go-aheads from your lock screen: paste one connector link, add standing instructions, and every session reaches you anywhere.

AG
Aadil Ghani
Founder, Pushary
Jul 10, 20263 min read
Share

To approve Claude Cowork tasks from your phone, connect Pushary as a custom connector (one pasted link, works on every paid Claude plan), then add standing instructions so Cowork asks before risky steps and notifies you when tasks finish. Questions arrive as push notifications, you answer yes or no from the lock screen, and the answer flows back into the running session.

Key takeaways

  • Setup is a paste, not an install: one connector link in Claude Settings, one standing-instructions block in Claude Settings, Cowork.
  • Answers happen on a signed decision page from the lock screen. No Claude app needed on the answering device.
  • The same inbox covers Claude Code, Codex, Cursor, and Hermes, and every question and answer lands in an exportable audit trail.

Sign in at pushary.com and open Settings, Connections. Copy the OAuth connector URL: https://pushary.com/api/mcp/mcp. It contains no API key. You will sign in and authorize Pushary when Claude connects.

Step 2: paste it into Claude

In Claude, open Customize > Connectors > Add custom connector. Paste the URL, leave OAuth Client ID and Secret empty, and sign in with the same Pushary account as your phone and Mac app.

Custom connectors are configured at the account level and work across claude.ai, Claude Desktop, Cowork, and the mobile apps. Inside a Cowork session, open Customize, then Connectors, and turn Pushary on. The Pushary dashboard flips to Connected on the first handshake.

Step 3: add the standing instructions

This is the step people skip, and it is the one that makes the loop proactive. Cowork only calls a connector when something tells it to. The Connections page has a ready-made block, Make Cowork reach out proactively; copy it into Claude Settings, then Cowork, under standing instructions.

The instructions guide Cowork to:

  • ask for missing decisions and approval for risky actions outside your existing authorization, using confirm, select or input,
  • send a notification with a short summary when a task finishes or fails,
  • identify itself as Claude Cowork - <task name> so your fleet view groups sessions correctly.

Step 4: answer from the lock screen

When Cowork asks a question, Pushary routes it according to your delivery settings. Supported confirm notifications offer lock-screen actions; choices and text open the app. Answer a harmless setup question and verify that the still-running Cowork session receives the answer.

Two details worth knowing:

  • Questions have a bounded lifetime. Cowork follows the returned timeout and handoff. A pending phone question is canceled before the agent asks in the current chat; silence is never approval.
  • A reply needs a live question and session. An embedded question can collect an unresolved decision, but it does not guarantee an overnight wait or restart a finished turn. Reopen Cowork for follow-up work.

What this does and does not do

Honesty section. This Pushary connector is a cooperative integration: Claude decides when to ask, guided by your standing instructions. It is not an enforced gate, and this setup does not verify native permission hooks. If you want approvals that physically block a command until you answer, that exists for Claude Code, Codex, Gemini CLI, Cursor, and Hermes via npx @pushary/agent-hooks setup.

Also worth knowing: Anthropic has its own notification loop for Cowork through the Claude mobile app, rolling out across plans as of mid-2026. If Cowork is your only agent, it may be enough. The honest comparison covers when it is and when it is not.

Why route it through one inbox

The average person delegating to Cowork is also running something else: a Claude Code session on a side project, a Codex run, a Cursor agent. Each has its own notification story, which in practice means none of them reliably reach you. One connector and one hook setup put all of them in a single phone inbox with a single history, and every question and answer is kept in an exportable audit trail so you can see what was decided while you were in a meeting.

Full setup details live on the Claude Cowork page and in the docs guide.

Frequently asked questions

How do you approve Claude Cowork tasks from your phone?

Connect Pushary as a custom connector with one pasted link, then add a block of standing instructions so Cowork asks before risky steps and notifies you when tasks finish. Questions arrive as push notifications and you answer yes or no from the lock screen on a signed decision page, and the answer flows back into the running session.

Is the Claude Cowork phone approval an enforced gate?

No. This Pushary connector does not install native permission hooks, so this is a cooperative integration where Claude decides when to ask, guided by your standing instructions, and this setup does not verify native permission hooks. Approvals that physically block a command exist for Claude Code, Codex, Gemini CLI, Cursor, and Hermes via npx @pushary/agent-hooks setup.

What happens if you miss a Claude Cowork notification?

A question is answerable only while it remains live. Cowork follows the timeout and handoff returned by Pushary, canceling a pending phone question before moving it to the current chat. A late answer cannot restart an ended Cowork turn.

AG
Aadil Ghani
Founder, Pushary

Building Pushary so an AI agent can reach you on your phone and wait for a yes before it does something you would not want.

Read next

Guides

How to add human-in-the-loop to a Vercel AI SDK agent

The AI SDK gives your agent tools. One of those tools can be a real person. Here is the whole integration, from install to fail-closed decision.

Jul 17, 20262 min readAadil Ghani
Guides

LangGraph human-in-the-loop: interrupt, approve, and resume from a phone

LangGraph pauses the graph and persists state. It does not notify anyone. Here are the two patterns that finish the job, with the footguns written down.

Jul 17, 20263 min readAadil Ghani
Guides

An audit trail for Claude Cowork: what did your AI coworker decide while you were out?

Delegation without a record is hope. Keep every Cowork question and answer in an immutable, exportable trail, next to every other agent you run.

Jul 10, 20264 min readAadil Ghani

Get a push the moment your agent needs you

Approvals, done alerts, and a kill switch for Claude Code, Codex, Cursor, and the rest. It takes a couple of minutes to set up.