Mcp-GdbBLOCK
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
[](https://www.npmjs.com/package/mcp-gdb) [](https://github.com/signal-slot/mcp-gdb/blob/main/LICENSE)
A Model Context Protocol (MCP) server that provides GDB debugging functionality for use with Claude or other AI assistants.
Features
- Start and manage GDB debugging sessions
- Load programs and core dumps for analysis
- Set breakpoints, step through code, and examine memory
- View call stacks, variables, and registers
- View source code with VS Code integration
- Execute arbitrary GDB commands
Installation
Claude Code
claude mcp add gdb -- npx -y mcp-gdb
Claude Desktop
Add the following to your Claude Desktop MCP configuration:
{
"mcpServers": {
"gdb": {
"command": "npx",
"args": ["-y", "mcp-gdb"]
}
}
}Install from Source
git clone https://github.com/signal-slot/mcp-gdb.git cd mcp-gdb npm install npm run build
Usage
Example Commands
Here are some examples of using the GDB MCP server through Claude:
Starting a GDB session
Use gdb_start to start a new debugging session
Loading a program
Use gdb_load to load /path/to/my/program with the sessionId that was returned from gdb_start
Setting a breakpoint
Use gdb_set_breakpoint to set a breakpoint at main in the active GDB session
Running the program
Use gdb_continue to start execution
Examining variables
Use gdb_print to evaluate the expression "my_variable" in the current context
Getting a backtrace
Use gdb_backtrace to see the current call stack
Terminating the session
Use gdb_terminate to end the debugging session
Supported GDB Commands
gdb_start: Start a new GDB sessiongdb_load: Load a program into GDBgdb_command: Execute an arbitrary GDB commandgdb_terminate: Terminate a GDB
d219d597ecc4OBSERVED · 2026-10-07Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.
claude mcp add mcp-gdb -- npx -y [email protected]
{
"mcpServers": {
"mcp-gdb": {
"command": "npx",
"args": [
"-y",
"[email protected]"
]
}
}
}Exposed tools (19)
13 read · 5 write · 1 destructive. Blast radius: 1 tool can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.
| Tool | Risk | Description |
|---|---|---|
gdb_attach | read | Attach to a running process |
gdb_backtrace | read | Show call stack |
gdb_command | write | Execute a GDB command |
gdb_continue | read | Continue program execution |
gdb_examine | read | Examine memory |
gdb_finish | write | Execute until the current function returns |
gdb_info_registers | read | Display registers |
gdb_list_sessions | read | List all active GDB sessions |
gdb_list_source | read | List source code at current location or specified location, with VS Code integration |
gdb_load | read | Load a program into GDB |
gdb_load_core | read | Load a core dump file |
gdb_next | read | Step over function calls |
gdb_print | read | Print value of expression |
gdb_reverse_continue | read | Reverse continue program execution |
gdb_set_breakpoint | write | Set a breakpoint |
gdb_start | write | Start a new GDB session |
gdb_step | read | Step program execution |
gdb_terminate | destructive | Terminate a GDB session |
gdb_watch | write | Set a watchpoint on a memory address or expression |
Trust audit
BLOCKgrade D · trust 65/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | FAIL |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (1 observation(s))
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (7)
echo "On Ubuntu/Debian: sudo apt-get install gdb"
echo "On Fedora/RHEL: sudo dnf install gdb"
echo "On Arch Linux: sudo pacman -S gdb"
gdb_terminate
@modelcontextprotocol/sdk, @types/node, @typescript-eslint/eslint-plugin, @typescript-eslint/parser, eslint, tsc-watch, typescript
system use found in code, not declared in the description
Gates applied: no_behavioural_pass, no_license.
d219d597ecc4full audit observations/trust-audit/mcp-server/signal-slot__mcp-gdb.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-07 | d219d597ecc4 | BLOCK | D | 65 | first audit |
Questions
What tools does Mcp-Gdb expose?
19 in total: 13 read-only, 5 that write, and 1 that can delete or overwrite (gdb_terminate). Every one is listed on this page with its risk.
Is Mcp-Gdb safe to connect to an agent?
No — not without reading the findings first. The audit graded it D (65/100) and found 3 critical or high issues in the source. Each one is listed on this page with the file and line it is on. Separately from the audit: 1 of its tools can destroy data, so scope the token you give it to what you actually need.
What credentials does Mcp-Gdb need?
No credential environment variables were found in its source, so it appears to need none.
How does Mcp-Gdb run?
It speaks stdio, so it runs as a local process your client starts. It is published on npm as mcp-gdb at 0.1.3.
How current is this page?
The grade is for one exact copy of the source (d219d597ecc4), read on 2026-10-07. The repository is watched and re-audited when it changes.