C2Prog includes c2p-agent, an MCP server that exposes C2Prog to an AI
assistant running on the same machine. The assistant can, for example, search
the C2Prog manual, report how your TI C2000™ target is connected, or read
memory from a running device.
Model Context Protocol is an open standard for connecting AI assistants to local tools. The server is host-agnostic. Any MCP-compatible host can launch it, including Claude Code, Codex, Claude Desktop, Cursor, VS Code in GitHub Copilot agent mode, OpenCode and Gemini CLI.
Registering the server
In the C2Prog GUI, open Help > Set up AI agent integration…. The dialog lists the supported agent hosts. For each one it shows whether the host is installed and whether the server is already registered. Select the hosts you want and click Register.
The register subcommand of c2p-agent does the same thing from a shell:
c2p-agent register
With no option, it registers every supported host that is installed on the computer. A host counts as installed when its configuration directory exists. For Claude Code, Codex and OpenCode, the command also installs a skill that describes the tools.
To register individual hosts, name them with --host:
c2p-agent register --host cursor --host vscode
c2p-agent register --status lists every host with its configuration file,
whether it is installed, and whether its entry launches this c2p-agent or
is configured for another C2Prog installation.
A host that is already running keeps using the server it started. Therefore, restart the host, or reconnect its MCP servers, after registering. On Windows, the installer’s Set up AI agent integration option runs the registration for you.
Searching the manual
Searching the manual is the most broadly useful of these capabilities. The manual ships with C2Prog, and the search runs on your machine.
Ask the assistant a question such as:
Explain to me step by step how to key provision an F29x device.
The answer spans several manual sections: generating the Manufacturer keys,
building the key-provisioning certificate, reviewing its consequence report,
creating the Extended Hex file, and programming it. The assistant reads those
sections in full, follows their cross-references, and gives you the procedure in
order with the c2p-cli command lines.
The manual answers narrow questions too. Ask what a flash error code means,
which options a c2p-cli command takes, or what a function in the DLL API does.
Inspecting the setup
Questions can also be about the machine in front of you:
Tell me the serial number of the attached XDS emulators.
The assistant reports the serial ports and the connected TI JTAG probes, each
probe with its serial number. That serial number is what goes into the port
configuration when several probes of the same type are connected, as in
xds110:<sn>.
The assistant can also report the programmable devices with their --target
options, the configured TI EmuPack path, the active license tier, and the
integration signature of an .ehx file. JTAG
probe detection and JTAG targets require an EmuPack, so it can set that path for
you as well.
Reading and controlling a running target
For a target on a JTAG probe, the assistant works through C2Prog’s GDB server.
Retrieve the following memory range from the target and store it as an Intel Hex file.
The assistant reads the range and writes the records into a file. The read is non-intrusive, so the application on the target keeps running. Status and register reads work the same way.
Halting, resetting, stepping and writing change the state of the target. Your agent host can tell these apart from the reads and ask you before one runs.
Register the server with your agent host, and your AI assistant has C2Prog, its manual and your connected hardware within reach.