MCP Server for AI Assistants

A Claude Code session analyzing an embedded system with 3 FPGA’s communicating with each other and a USB host using the Active-Pro App with MCP Server

MCP Server for AI Assistants

The Active-Pro MCP Server lets AI assistants, including Claude, control the Active-Pro application directly. Once connected, you can ask the AI to start captures, configure triggers, search decoded data, export AI Snapshots, control outputs, navigate between recorded triggers, and run complete test sequences in plain English.

The server is a small Python script that sits between the AI assistant and the existing Automation API on port 37800. No changes to the Active-Pro application are required.

Download: active_pro_mcp.zip

The zip contains two files:

• server.py - the MCP server
• requirements.txt - the single Python dependency

What Is MCP?

MCP stands for Model Context Protocol - an open standard developed by Anthropic that lets AI assistants connect to external tools and services. Think of it as a plug-in system for AI: instead of just answering questions from memory, the AI can reach out to real applications and control them.

When you set up the Active-Pro MCP server, Claude gains a set of tools it can call, start_capture, search, export_ai_snapshot, set_trigger_mode, next_trigger, and others. When you ask Claude to "run a 10-second capture, trigger on the word FAULT, then walk me through every match," it doesn't generate instructions for you to follow manually; it actually calls those tools itself and reports back what it found.

You need Claude Desktop to use MCP servers. Claude Desktop is Anthropic's native application for Windows and Mac. The browser version of Claude at claude.ai does not support local MCP servers. Claude Desktop is free to download and includes the same AI as the web version.

Overview of the Setup Process

There are five things to do, in order:

1. Install Claude Desktop
2. Install Python
3. Install the MCP Python package
4. Extract the Active-Pro MCP server files
5. Tell Claude Desktop where to find the server

Total time: about 10-15 minutes.

Step 1, Install Claude Desktop

If you already have Claude Desktop installed and signed in, skip to Step 2.

1. Go to claude.com/download
2. Click Download for Windows (or the macOS version if you are on a Mac)
3. Run the installer and follow the prompts, accept the defaults
4. Launch Claude Desktop from the Start menu
5. Sign in with your Anthropic account (the same account you use at claude.ai, or create a free one if you don't have one yet)

Important: Always download Claude Desktop directly from claude.com/download. Do not use third-party download sites or the Microsoft Store version, they may not support MCP.

You can confirm Claude Desktop is working by opening it and having a quick conversation. It works exactly like the web version.

---

Step 2, Install Python

The MCP server is a Python script, so Python must be installed on your PC. This is a different Python than the one bundled inside the Active-Pro application, the bundled interpreter is used only to run Custom Decoders inside the app, and is not available to Claude Desktop.

1. Go to python.org/downloads
2. Click the large Download Python button, you want version 3.10 or above (the site shows the current latest, which is fine)
3. Run the installer
4. On the very first screen of the installer, check the box that says "Add Python to PATH" - this is the most common mistake people make. If you miss it, Python installs but won't be usable from the command line.

To confirm Python installed correctly: Open a Command Prompt (press the Windows key, type cmd, press Enter) and type:

python --version

You should see a response like Python 3.12.3. If you see an error like 'python' is not recognized as an internal or external command, Python is not on your PATH, uninstall it and reinstall with the "Add Python to PATH" checkbox checked.

Step 3, Install the MCP Package

The server needs one Python package. Install it with a single command.

1. Open Command Prompt or PowerShell (press the Windows key, type cmd, press Enter)
2. Type the following exactly and press Enter:

pip install mcp

You will see progress output as it downloads and installs. When it finishes and returns you to a prompt (>), it is done. The whole thing takes about 10-30 seconds depending on your connection.

If you see "pip is not recognized": Try this alternative command instead:

python -m pip install mcp

If that also fails, Python's PATH setup did not complete, reinstall Python with the "Add to PATH" box checked.

If you see a permissions error on Windows: Close the Command Prompt, then right-click Command Prompt in the Start menu and choose Run as administrator. Try the pip install mcp command again.

---

Step 4, Extract the MCP Server Files

1. Download active_pro_mcp.zip from the link above
2. Extract (unzip) the contents to a permanent location on your PC, somewhere you won't accidentally move or delete. A good path is:

C:\ActivePro\mcp\

After extracting, confirm these two files exist at that location:

C:\ActivePro\mcp\server.py
C:\ActivePro\mcp\requirements.txt

Leave these files in place permanently. Claude Desktop launches server.py as a background process each time it starts. If you move the file later, Claude Desktop will fail silently and the hammer icon won't appear. If you ever move the files, update the configuration path described in Step 5.

Step 5, Configure Claude Desktop

Claude Desktop reads a JSON configuration file to find out which MCP servers to connect to. You need to add an entry for the Active-Pro server.

Finding the Configuration File

The configuration file lives at:

%APPDATA%\Claude\claude_desktop_config.json

The quickest way to get there:

1. Press Windows key + R to open the Run dialog
2. Type %APPDATA%\Claude and press Enter
3. Windows Explorer opens the Claude application data folder

Look for a file called claude_desktop_config.json.

If the file does not exist yet: That is normal for a fresh installation. Create it: right-click in the Explorer window, choose New > Text Document, and name it claude_desktop_config.json. Make sure the file name ends in .json and not .json.txt.

Tip: If you cannot see file extensions in Windows Explorer, click the View tab in Explorer's ribbon and check File name extensions.

Alternatively, you can open the config directly from within Claude Desktop: go to Settings (gear icon) > Developer > Edit Config.

Writing the Configuration

Open claude_desktop_config.json in any text editor, Notepad works fine. The complete file contents should look exactly like this:

{
    "mcpServers": {
    "active-pro": {
    "command": "python",
    "args": [
    "C:\\ActivePro\\mcp\\server.py"
    ]
    }
    }
}

Critical details:

• Use double backslashes (\\) in the Windows path. A single backslash is a special character in JSON and will cause a silent failure.
• The file must be valid JSON. The most common mistakes are: a missing comma between entries, a trailing comma after the last entry, mismatched braces, or an unclosed quote.

If you already have other MCP servers listed in the file, don't replace the whole file, add the "active-pro" entry inside the existing "mcpServers" block:

{
    "mcpServers": {
    "some-other-server": {
    "command": "node",
    "args": ["C:\\path\\to\\other\\server.js"]
    },
    "active-pro": {
    "command": "python",
    "args": [
    "C:\\ActivePro\\mcp\\server.py"
    ]
    }
    }
}

Save the file.

Step 6, Restart Claude Desktop and Verify

Fully quit Claude Desktop - this is important. Simply closing the window often leaves Claude running in the background (you can see it in the system tray at the bottom-right of the taskbar). To actually quit:

• Right-click the Claude icon in the system tray and choose Quit, or
• Go to File > Exit in the Claude Desktop menu bar

Then relaunch Claude Desktop from the Start menu and open a new conversation.

Look for the hammer icon (🔨) in the bottom-left of the chat input bar. This icon means Claude has MCP tools available. If it is there, setup is complete.

Click the hammer icon to expand the tool list. You should see the Active-Pro tools, start_capture, stop_capture, search, the trigger tools, and the rest. If they are listed, everything is working.

If the hammer icon does not appear, see the Troubleshooting section at the bottom of this page.

Step 7, Your First Commands

Before asking Claude to control the Active-Pro:

• Make sure the Active-Pro application is open and running (version 3.8 or above, 6.x recommended for trigger commands)
• Connect your hardware device if you plan to take captures

Start with a simple verification:

"Are you connected to the Active-Pro? Is a device plugged in?"

Claude will call the is_connected tool and report back. If it says yes, you are ready to go. If it says no, check that the Active-Pro application is running.

Example Prompts to Try

Basic capture:

"Start a 5-second capture, wait for it to finish, then tell me how many bytes were captured."

Search for problems:

"Run a 10-second capture. When it's done, search for any message containing ERROR. If you find one, tell me the timestamp and save the capture to C:/captures/debug.active."

Triggered capture:

"Configure a trigger that fires when Device A text contains FAULT, with 2 seconds of pre-trigger context. Stop after 5 triggers. Start the capture and walk me through each captured fault."

Stimulus and capture:

"Set A1 to a ramp waveform from 0 to 3.3V, capture for 3 seconds, then zoom to show all captured data and save a screenshot to C:/docs/ramp_test.png."

Full AI analysis loop:

"Start a 5-second capture. When it finishes, export an AI Snapshot for the full time range to C:/temp/snap.aft. Open that file and tell me what you observe about the firmware output."

Automated pass/fail test:

"Run a 10-second capture. If the word ERROR appears anywhere in the decoded output, save the capture to C:/failures/ with today's date in the filename and explain what happened. Otherwise just report pass."

Behind the scenes, Claude is calling the same Automation API commands described in the Automation API chapter, it just decides which commands to call, in what order, based on your instructions.

Multiple Active-Pro Instances

When more than one Active-Pro instance is running simultaneously, each uses a different port (37800, 37801, 37802, ...). Every MCP tool accepts an optional device parameter (1-8, default 1) that selects which instance to target.

Example: "Start a capture on device 2" tells Claude to target the instance listening on port 37801.

---

Available Tools

The MCP server exposes the Automation API as individual typed tools. The 96 tools below cover capture control, signal reading, output control, trigger configuration and navigation, buffer settings, file operations, view control, and notes / analysis context.

For anything that does not have a dedicated tool, use send_raw_command to pass a raw API command string. See the Automation API chapter for the full raw command set.

Connection and status

Tool		Purpose
hello		Liveness check.
is_connected		Returns YES/NO.
is_capturing		Returns YES/NO.
get_devices_attached		Returns USB serial numbers of attached pods.
select_device		Switch the active pod by index.

Capture control

Tool		Purpose
start_capture		Begin a capture.
stop_capture		Stop and complete post-processing.
new_capture		Clear the current capture buffer.
get_capture_size		Current capture size (bytes).
get_capture_time		Capture duration in seconds.

Signal reading (requires active capture)

Tool		Purpose
get_logic		8-bit bitmask of D0-D7.
get_analog		Voltage on analog channel 1-6 (4-6 are Ultra-only).

Analog channel settings (read-only)

Tool		Purpose
get_channel_offset		Per-channel offset (channel 1-8).
get_channel_scale		Per-channel scale multiplier.
get_channel_units		Per-channel units label.
get_current_resistor		Current-measuring resistor value in ohms.
get_battery_capacity		Battery capacity in mAh.
get_analog_range		0_20 or MINUS10_PLUS10.
get_analog_sample_rate		Configured analog sample rate (samples/sec).

Digital output control

Tool		Purpose
set_digital_output_mode		D0/D1 output mode (tri-state, low, high, PWM).
set_digital_pwm		PWM duty 0-255 on D0/D1.

Analog output control

Tool		Purpose
set_analog_output_mode		A0/A1 mode (tri-state, fixed voltages, custom DC, waveforms on A1).
set_analog_dc_level		Custom DC voltage for A0/A1 when mode = 6.
set_a1_waveform		A1 ramp/sine/square/triangle min, max, step count.

Cursor control

Tool		Purpose
set_cursor		Move the current, x1, or x2 cursor to a time (seconds).
get_cursor		Return current, x1, or x2 cursor time.

Post-capture pipeline

Tool		Purpose
apply_changes		Re-run Custom Decoders, PacketPresenters, then trigger search. Blocks until done or until the GUI Cancel is clicked.
is_processing		Non-blocking check of pipeline state.
cancel_processing		Request cancellation of the running pipeline.

Trigger status and navigation

Tool		Purpose
is_triggered		YES if any triggers fired.
get_trigger_count		Total triggers detected.
get_trigger_index		1-based index of the currently viewed trigger.
get_trigger_cursor		Trigger cursor position in seconds.
set_trigger_cursor		Move the trigger cursor to a time.
next_trigger, prev_trigger		Step the trigger cursor.
first_trigger, last_trigger		Jump to the oldest or most recent trigger.
goto_trigger		Jump to trigger N by 1-based absolute index.
center_trigger		Re-center the current trigger in the waveform.
reprocess_triggers		Re-scan with current trigger settings only (legacy, prefer apply_changes).

Trigger configuration

Tool		Purpose
set_trigger_mode, get_trigger_mode		off / normal / auto.
set_trigger_source_type, get_trigger_source_type		none, analog, digital, atext, avalues, ..., dppgraph.
set_trigger_channel, get_trigger_channel, get_trigger_channel_list		Pick the channel within the current source.
set_trigger_condition, get_trigger_condition		rising / falling / pulsehighgreater / pulsehighless / pulselowgreater / pulselowless / contains / equals.
set_trigger_threshold, get_trigger_threshold		Analog / Device-Values threshold value.
set_trigger_pulse_width, get_trigger_pulse_width		Value plus unit (ns/us/ms/s).
set_trigger_text, get_trigger_text		Pattern for contains / equals.

Buffer and pre / post-trigger

Tool		Purpose
set_buffer_size, get_buffer_size		Capture buffer size in MB.
set_pre_trigger_mode, get_pre_trigger_mode		keepall / keeplast.
set_pre_trigger_seconds, get_pre_trigger_seconds		Keep-last window depth (1-60 s).
set_post_trigger_mode, get_post_trigger_mode		untilfull / stopafterseconds / stopaftertriggers.
set_post_trigger_seconds, get_post_trigger_seconds		Seconds to keep capturing after first trigger.
set_post_trigger_triggers, get_post_trigger_triggers		Number of triggers to capture before stopping.

Input channel settings

Tool		Purpose
logic_channel_on_off		Enable / disable a logic channel D0-D7.
analog_channel_on_off		Enable / disable an analog channel CH1-CH8.
set_device_mode		Set a device port (A-D) decoder mode by decoder index.
set_threshold_level		Digital logic threshold (1.0, 1.8, 2.5, 3.3, 5.0 V).
set_label		Set the display label for a logic, analog, or device entity.
set_device_setting		Set the numeric setting (baud, bit rate) for a device port.

View and navigation

Tool		Purpose
zoom_all		Fit the whole capture on screen.
zoom_from		Zoom to a time range.
search		Search forward for text in decoded ADP data.

Panels and display

Tool		Purpose
show_tab		Open a tab: inputs, triggers, outputs, list, settings, notes, code, liveui, devices, close.
set_device_visibility		Show / hide a device's rows in the waveform.
set_theme		Switch dark or light mode.

Notes

Tool		Purpose
clear_note		Wipe the Notes editor.
append_note		Append text.
get_note		Read the full Notes content.

AI Snapshot Analysis Context

Tool		Purpose
clear_analysis_context		Wipe the Analysis Context.
append_analysis_context		Add a line to the context block written into AI Snapshot exports.
get_analysis_context		Read the current Analysis Context.

File operations

Tool		Purpose
open_capture, save_capture		.active capture file load / save.
save_between_cursors		Save the X1-X2 range to a new .active file.
open_configuration, save_configuration		Configuration .ini load / save.
export_between_cursors		Export decoded data as CSV.
export_ai_snapshot		Export the X1-X2 range as a .aft AI Snapshot.
save_screenshot		Save a PNG of the full application window.

Application

Tool		Purpose
exit_app		Close the application.
send_raw_command		Pass-through to the TCP API, for anything not covered by the typed tools.

Custom Decoders are configured via the GUI, not the MCP server. There are no MCP tools to attach a Custom Decoder, set its parameters, or edit its script. The intended workflow is to set up your decoders (and channel labels, units, triggers, etc.) interactively in the GUI, save a configuration with save_configuration, then in scripts call open_configuration <path> to apply it, followed by apply_changes to rerun the pipeline if a capture is already loaded. See the Custom Decoders chapter for the GUI-side setup and the recommended save / restore workflow.

macOS Differences

The steps above are written for Windows. On macOS the differences are:

Python: Open Terminal and type python3 --version. If it prints 3.10 or above, you are set. Otherwise download from python.org.

Install the dependency:

pip3 install mcp

Server file location: Extract to somewhere like /Users/yourname/ActivePro/mcp/ - use forward slashes.

Config file location:

~/Library/Application Support/Claude/claude_desktop_config.json

To open that folder in Finder: press Cmd+Shift+G, paste the path above, press Enter.

Config file content on Mac - use python3 and forward slashes:

{
    "mcpServers": {
    "active-pro": {
    "command": "python3",
    "args": [
    "/Users/yourname/ActivePro/mcp/server.py"
    ]
    }
    }
}

Use whichever command (python or python3) works in your Terminal.

Troubleshooting

The hammer icon does not appear after restarting Claude Desktop

Work through these checks in order:

1. Verify the JSON is valid. Open claude_desktop_config.json in a text editor and look for: mismatched braces, missing commas between entries, a trailing comma after the last entry, or unmatched quotes. A single character mistake silently prevents the config from loading.

To check automatically: copy the entire file contents and paste into an online JSON validator such as jsonlint.com. It highlights errors with the line number.

2. Verify the path to server.py is correct. Open Command Prompt and type exactly what you have in the config (as a single command):

python C:\ActivePro\mcp\server.py

If you see No such file or directory or any path error, the path in your config is wrong. Find the actual location of server.py and update the config.

3. Verify Python is on your PATH. Type python --version in Command Prompt. If it fails, Python is not accessible to Claude Desktop. Reinstall Python and check "Add to PATH."

4. Confirm Claude Desktop was fully quit and restarted. On Windows, closing the window often leaves Claude in the system tray. Right-click the system tray icon and choose Quit, then relaunch.

5. Read the Claude Desktop logs. Claude Desktop writes MCP error details to log files:

• Windows: %APPDATA%\Claude\logs\
• macOS: ~/Library/Logs/Claude/

Open the most recent log file and search for active-pro or error. The message will usually pinpoint the problem.

"Cannot connect to Active-Pro on port 37800"

The Active-Pro application is not running, or the installed version is below 3.8. Start the Active-Pro application first, then ask Claude to try again.

Commands time out

The default timeout is 15 seconds. stop_capture on very long captures with heavy decoding (especially with several Custom Decoders attached) can take longer, this is expected. The server waits for the Active-Pro application to finish processing before returning. If timeouts happen consistently, try shorter capture durations or reducing the number of active channels.

Claude says it doesn't have any Active-Pro tools

You are either using the browser version at claude.ai (which does not support local MCP servers), or Claude Desktop did not successfully load the config. Confirm the hammer icon appears in Claude Desktop's chat input bar, and that you started a new conversation after restarting.

pip install mcp fails

Try:

python -m pip install mcp

On macOS:

pip3 install mcp

On Windows, if you see a permissions error, open Command Prompt as Administrator and try again.

Quick-Start Checklist

Use this before asking Claude to control the Active-Pro for the first time:

• Claude Desktop installed from claude.com/download and signed in
• Python 3.10 or above installed, python --version works in Command Prompt
• pip install mcp completed without errors
• server.py extracted to a permanent folder (e.g. C:\ActivePro\mcp\)
• claude_desktop_config.json saved with correct path and double backslashes
• Claude Desktop fully quit (not just window closed) and relaunched
• Hammer icon (🔨) visible in the Claude Desktop chat input bar
• Active-Pro application running (version 3.8 or above; 6.x for trigger commands)
• Hardware device connected (if doing captures)

Import the included ACTIVE_PRO_MCP.md file into the tools base instruction file

Telling Claude to allow the Active Pro commands (otherwise you’ll be prompted the first time you use any of them)