Files, with context
Find a file, read its contents, and make a precise edit. Work inside the directories you’ve allowed.
read_file · edit_block · start_searchBetween your AI and your devices
AI works on your computer. Uses fewer tokens.
Connect ChatGPT, Claude, Cursor and other remote MCP-enabled AI apps to your PCs and servers. Find files, run code and review results. AI Efficient Mode helps reduce unnecessary tool output and AI input tokens.
Connect a computer first
Then, connect your AI app
ChatGPT · Claude · Cursor · other remote MCP apps
Official ChatGPT and Claude listing links are coming soon.Other MCP client setup →
Trading-PCConnectedInteractive demo
Result ready
- Understand the requestIdentify the files and tools for the task.
- Run on the selected deviceUse the permissions you have allowed.
- Return the resultReview the output and continue.
start_process/projects/app
$ python -m pytestcollected 12 items12 passed · exit code 0
Illustrative demo · Runs with your permissions.
Your PCs and servers,
all in one place.
AI directs each device through RemoteCommand, interprets its results, and uses them to guide work on another device.
For example, inspect code on a PC, then ask a server to run the relevant tests. AI coordinates the steps; devices do not collaborate autonomously.
- Set access boundaries per device
- Review activity in one place
ChatGPT · ClaudeCursor · Other MCP
RemoteCommand
- PC
- Laptop
- Servers
Files · TerminalGit · Docker
Less repetitive work.
More room to focus.
Keep your tools and files where they are. Bring the work into the conversation.
Commands that keep going
Start a process, follow its output, and return to the result. Long-running work stays visible in sessions.
start_process · read_process_outputYour development workflow
Inspect Git changes and work with Docker containers using the tools on your PCs and servers.
git_status · git_diff · docker_logsSMARTER OUTPUT, NOT LESS CONTEXT
Give AI the signal. Not the noise.
RemoteCommand's AI Efficient Mode reduces repetitive command output, keeps requested files accessible, and lets AI continue reading long results in ranges.
How it works
See the differenceAnimated illustration · not a benchmark
Repeated progress messages make it harder to spot what matters.
[09:41:03] DEBUG heartbeat...
[09:41:04] INFO build 42/100
[09:41:05] DEBUG heartbeat...
[09:41:06] WARN retry attempt 2
[09:41:07] DEBUG heartbeat...
[09:41:08] ERROR test suite failed
[09:41:09] DEBUG heartbeat...
Reduce repetitive noiseAI Efficient Mode
WARN retry attempt 2
ERROR test suite failed
Repeated progress lines summarized
Less log noise, clearer next stepsChat with the important result instead of pages of repetitive logs.
Less repetitive output
Get to useful signals faster when logs are noisy.
Requested source preserved
Read exact code and documents when the task needs them.
Continue long results
Explore more output in ranges instead of losing context.
Reducing tool output can save AI input tokens. Your dashboard measures output data reduction in bytes and, for supported agent runs, estimates saved input tokens from actual text. These metrics do not directly measure your AI bill; results vary by task.
Connect in three steps.
Setup guide01
Sign in
Create your RemoteCommand account.
02
Register the agent
Install the agent on each PC or server and set its file and command permissions.
03
Connect your AI
Connect MCP, review permissions, and start your first task.
MCP connection address+
https://api.remotecommand.app/mcpChoose your RemoteCommand workspace
Start with a monthly allowance. Move to Pro when your projects grow.
Free
$0/ month
Your own AI-to-computer workspace, ready to explore.
- 10,000 requests per month
- AI Efficient Mode filters out noisy tool output
- Access your registered computers from AI
Pro
US$20/ month
For bigger workloads that should not stop at a monthly quota.
- No monthly tool-call cap
- AI Efficient Mode keeps tool responses focused
- Keep working across your connected computers
- Priority Support
Works with ChatGPT, Claude, and other MCP-compatible apps. Local taxes and final pricing are shown before payment.
Let AI handle the task.
You set the boundaries.
Allowed folders · Command approval · Activity logs · Revoke access
Before you connect
How does work span multiple devices?+
AI chooses a device for each task through RemoteCommand, reviews the returned output, and directs the next task to the appropriate device. Devices do not independently communicate or delegate tasks to one another. Each task follows the target device's permissions and approval policies.
Does RemoteCommand control my screen?+
RemoteCommand connects AI tools to files, terminal commands, Git, Docker, and processes. It does not currently stream your screen or operate your mouse.
Where do my files go?+
Your working files remain on each device. Requested file content and command output travel through the RemoteCommand relay to the AI service you connect. Review that AI service’s data policies too.
Can I use ChatGPT and Claude?+
RemoteCommand provides a remote MCP endpoint with OAuth. Connect it from an AI client that supports remote MCP. Available connector settings depend on your AI client and account; this is not an official OpenAI or Anthropic product.
Which devices can I connect?+
RemoteCommand connects Windows, macOS and Linux computers with one Node.js agent. Install Node.js LTS, then run npx -y @lhtsoft/remotecommand-agent@latest in a terminal. No separate agent installer is needed.
How do I update or fix a missing npx command?+
Stop the running agent with Ctrl+C, then run npx -y @lhtsoft/remotecommand-agent@latest run as the same OS user. If npx is missing, install Node.js LTS from nodejs.org and reopen your terminal. Background services are updated separately by their administrator.
Are terminal commands limited to allowed folders?+
Allowed directories constrain file tools. Approved terminal commands run automatically with the agent user's operating system permissions and may affect files outside those folders. Enable only the capabilities you need.
Why might ChatGPT or Claude block a tool request during safety or permission checks?+
ChatGPT and Claude use separate safety checks and tool permission controls, so their warning messages and decisions may differ. A request may be blocked before it reaches RemoteCommand; this does not necessarily mean a RemoteCommand outage or a monthly usage limit. Check the requested action and permissions granted to your AI client. For file or Git inspection, prefer read-only tools such as read_file or git_status instead of execute_code. Do not try to bypass a block on an unsafe action. If a legitimate request is repeatedly blocked, contact support with the AI client name, tool name, approximate time, and exact error message. RemoteCommand cannot disable either provider's safeguards.