ChannelTerm CLI
A shared terminal for a human and an AI: one serial or SSH session, board output reading, GDB, file transfer and an MCP server
- 145 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- The agent sends commands into a real device and can break board settings or firmware
- The shared session exposes SSH and file transfer to the device, so an isolated network and limited rights are needed
Install
- Works in
Manual install
Read the README of https://github.com/akira-init1/ChannelTerm and `docs/getting-started/mcp-server.md`. Download the archive from Releases only after the user agrees and verify it against `SHA256SUMS`. Send commands to the device on the user request and wait for confirmation before dangerous actions.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent sends commands into a real device and can break board settings or firmware
- The shared session exposes SSH and file transfer to the device, so an isolated network and limited rights are needed
This is third-party code. Review the repository files before installing.
What it does
ChannelTerm gives a human and an AI client a shared live terminal: a serial port or SSH. Both read the same device output, see each other commands and take turns operating the board. The program runs on Windows, Linux and macOS, and external AI clients connect through MCP. Each client reads independently, so the AI does not consume the output you are watching, and its commands appear in your window. With GDB the agent sets breakpoints, steps through Linux programs and inspects variables: the serial port is connected on Windows and GDB runs in WSL2, or the shared console stays open while GDB goes over the board Ethernet. The host can interrupt U-Boot, verify the prompt and export a diagnostic ZIP, optionally with raw bytes. Files are transferred over the shared serial session with SHA-256 checks or over dedicated SFTP and SCP connections. The `channelterm attach` command can create the local shared host automatically.
Who it is for. Embedded developers and board debugging engineers who work with a serial port, U-Boot and GDB and want to attach an agent to the same terminal.
Good fit when
- A board is debugged over a serial port and the agent must read the same output
- You need breakpoints and variable inspection in a Linux program through GDB
- A short boot window requires interrupting U-Boot and confirming the prompt
- Files must be sent to the board with SHA-256 checks or over SFTP and SCP
Not a fit when
- There is no device to debug: a live board, serial port or SSH access is needed
- An agent must not be trusted with commands on hardware without you present
- You need only a plain SSH client with no AI
Example request
Attach to the shared serial session of the board, wait for the U-Boot prompt and show the last forty lines of boot outputLimitations
Ready archives for Windows, Linux and macOS come from the Releases section and are checked against `SHA256SUMS`, then `channelterm install` is run. Building from source needs Go 1.26 or newer. For serial debugging on Windows GDB runs in WSL2. The agent sends commands into the real device terminal and can break settings or firmware, so limit access beforehand. The last commit is dated October 24, 2026.
How to disable. Remove the ChannelTerm entry from the AI client MCP settings, stop the shared host and uninstall the program that `channelterm install` set up.
FAQ
What is a ChannelTerm shared session?
One live serial or SSH terminal that you and an AI client are both attached to: both read device output and see each other commands.
How do I connect an AI client?
Through MCP, and the `channelterm attach` command can create the local shared host automatically. Details are in `docs/getting-started/mcp-server.md`.
Can I debug through GDB?
Yes, the agent sets breakpoints, steps through the program and inspects variables. On Windows GDB runs in WSL2, or you work over the board Ethernet.
How are files transferred?
Over the shared serial session with SHA-256 checks or over dedicated SFTP and SCP connections.
Which operating systems are supported?
Windows, Linux and macOS, with archives in Releases.
Discussion
No questions or reviews yet.
Related
CLI toolsA Rust terminal with a multiplexer: agent sessions from several machines in one window, plus a skill for seeing adjacent panes
TMX is a desktop SSH client with terminals, SFTP and a built-in MCP server through which an agent drives sessions.
Console Automation MCP gives an agent local shells and SSH sessions: input, output, error detection, long processes
An MCP server giving ChatGPT and agents a terminal, files, Playwright and remote machines inside an isolated workspace.
Your machine's terminal, windows and desktop in a browser, plus an MCP server for agents to read and type into other sessions
A Tabby Terminal plugin that exposes terminal tabs to MCP clients: the agent sees sessions, runs commands and reads the buffer.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| ChannelTerm (this card) | no | not required | 145 | |
| SSH MCP | no | API key | 875 | a shared session for human and agent, plus serial port, GDB and U-Boot |
| Embedded Skills | no | 797 | a program for a live board terminal and not a set of skills |
README in short
The README presents ChannelTerm as a tool for debugging a board together with an AI in one terminal. It shows a shared session diagram, describes features (shared console, GDB, file transfer, automatic host) and installation from Releases archives or a Go build. Documentation is translated into nine languages.