Dock Agent

Your Claude Code sessions, one glance away — a floating macOS overlay for monitoring AI agents

Code File Edit View Wed 9 Apr 3:42 PM

☰

⊟

⌕

⎇

▶

TS handler.ts

TS middleware.ts

{ } types.ts

1import { Request, Response } from 'express'
2import jwt from 'jsonwebtoken'
3import { AuthError } from './types'
4
5const SECRET = process.env.JWT_SECRET
6
7/**
8 * Verify the incoming request token
9 */
10export async function verifyToken(req: Request) {
11  const token = req.headers.authorization
12  if (!token) throw new AuthError()
13  return jwt.verify(token, SECRET)
14}
15
16export async function refreshToken(req: Request) {
17  const decoded = await verifyToken(req)

mainTypeScriptLn 11, Col 42UTF-8

CC 2 active1

Dock Agent2 sessions

⌃

agent-notch

ClaudeVS Code2m

READsrc/main.jsx

EDITsrc/main.jsx

Bash Command rm -rf

                  rm -rf node_modules && npm install
                

in ~/Documents/agent-notch

my-saas-app

ClaudeiTerm5m

BASHnpm test -- --watch

EDITsrc/api/routes.ts

routes.ts — my-saas — Visual Studio Code

─

□

☰

⊟

⌕

⎇

▶

TS routes.ts

TS schema.ts

1import { Router } from 'express'
2import { authenticate } from './middleware'
3import { db } from '../db/schema'
4
5const router = Router()
6
7router.get('/users', authenticate, async (req, res) => {
8  const users = await db.query('SELECT * FROM users')
9  res.json(users)
10})
11
12router.post('/users', authenticate, async (req, res) => {
13  const { name, email } = req.body
14  const user = await db.insert(name, email)

mainTypeScriptLn 8, Col 30UTF-8

C my-saas1

Dock Agent1 session

⌃

my-saas-app

ClaudeTerminal4m

EDITsrc\api\routes.ts

READsrc\db\schema.ts

Bash Command git push --force

                  git push --force origin main
                

in C:\Users\dev\my-saas-app

⊞

▪ VS Code

▪ Terminal

3:42 PM

Everything you need

Monitor, approve, and control your Claude Code sessions from one place

🤖

Session monitoring

Track every Claude Code session running across your terminals and VS Code in a single floating overlay.

Claude Code more coming

🔒

Bash command approval

Review every bash command before it runs. Dangerous patterns like rm -rf and git push --force are highlighted in red.

⌨️

Keyboard-first

Approve, deny, and answer questions without leaving your keyboard.

⌘Y Approve

⌘N Deny

⌘1-9 Answer

📊

Rate limit tracking

See your 5-hour and weekly usage percentages at a glance. Know when you're approaching limits before you hit them.

🛡️

100% local

All communication stays on your machine via local IPC. No servers, no cloud, no telemetry.

📡

Real-time event feed

Live timeline showing every read, write, and command. Token counts and estimated costs per session.

Built for AI-native developers

If you're running Claude Code sessions all day, Dock Agent keeps you in control

💻

Solo developers

Running multiple Claude Code sessions across VS Code, Cursor, and Terminal? Dock Agent shows them all in one floating overlay so you never lose track of what your AI agents are doing.

🏢

Engineering teams

Give your team confidence when letting Claude Code touch production code. Every bash command is reviewed before execution, and dangerous patterns are flagged automatically.

🔬

Security-conscious users

Claude Code approval without YOLO mode. Review every shell command, file write, and network call before it executes — all locally on your Mac, no telemetry.

Frequently asked questions

Everything you need to know about Dock Agent

What is Dock Agent?

Dock Agent is a floating desktop overlay that monitors your Claude Code sessions in real time. It lets you review and approve every bash command, file edit, and tool call before Claude Code runs it — so you stay in control of your AI coding agent. Available on macOS and Windows.

How is this different from Claude Code's built-in approval?

Claude Code's built-in approval requires you to focus on each terminal window. Dock Agent unifies approvals from every session — across multiple terminals, VS Code, and Cursor — into one always-visible overlay. You can approve or deny with keyboard shortcuts without switching apps.

Does Dock Agent send my code or prompts anywhere?

No. All communication happens locally on your machine via inter-process communication. Dock Agent has no analytics, no telemetry, and no cloud backend. Your code, prompts, and session data never leave your machine.

Which AI coding tools does Dock Agent support?

Dock Agent currently supports Claude Code (Anthropic's official CLI and IDE extensions). Support for additional AI agents and coding tools is planned.

Which platforms does Dock Agent support?

Dock Agent is available on macOS (Apple Silicon and Intel) and Windows (x64). Linux support is planned.

How does the free trial work?

Download Dock Agent and get 3 days of full access — no credit card required. After the trial, a one-time $14.99 purchase activates the app permanently on one device. No subscription, no recurring charges.

macOS says the app is "damaged" or "cannot be verified"

This happens because the app is not yet notarized with Apple. Open Terminal and run: xattr -cr "/Applications/Dock Agent.app" then open the app normally. This only needs to be done once. A fully notarized build is coming soon.

Windows says "unknown publisher" or blocks the installer

When Windows SmartScreen shows "Windows protected your PC", click More info, then click Run anyway. This happens because the app is not yet code signed with an EV certificate. A signed build is coming soon.

Can I track rate limits and API usage?

Yes. Dock Agent shows real-time token counts and estimated API costs per session, plus a live timeline of every read, write, and command. You'll always know how close you are to your Claude rate limits.