The debugger for where IDEs can't go.

mcp-debugger is a headless MCP server that lets AI agents drive real debuggers — breakpoints, stepping, live variables — in CI runners, containers, Kubernetes pods, and cloud sandboxes. Seven languages. No IDE anywhere in the loop.

$ npx @debugmcp/mcp-debugger stdio
agent session — checkout pod, stagingpaused
  1. create_debug_session { language: "python" } ✓ session created
  2. attach_to_process { port: 5678 } ✓ attached to pod
  3. set_breakpoint { file: "app.py", line: 51 } ● verified
  4. # one request comes in …
  5. ■ stopped — "breakpoint" app.py:51 total = sum(prices)
  6. get_local_variables prices = [270] # expected [300]
  7. evaluate_expression { "_cart_cache" } {('cherry',): [270]}
  8. ✓ root cause: bulk discount mutates the shared cached list
  9. continue_execution ✓ pod kept serving throughout

where IDEs can't go

Debug the processes nobody can click on

An IDE debugger needs an IDE, a screen, and a human at the keyboard. Agents work where none of those exist — and that's exactly where the interesting failures live.

CI that explains itself

On test failure, the agent sets real breakpoints in the failing path, inspects live values, and posts the root cause as a PR comment.

- uses: debugmcp/mcp-debugger/.github/actions/debug-failing-test@main
The debug-failing-test action →

Sick pods, diagnosed live

Attach to a misbehaving pod through a port-forward, inspect the state your logs never captured, detach. No redeploy, no new logging.

$ kubectl port-forward deploy/checkout 5678:5678
Just-in-time diagnostics tutorial →

Cloud agents, real debuggers

Claude Code, Copilot, Cursor, and headless harnesses connect over stdio or HTTP. A debug session becomes structured tool calls an agent can reason about.

$ claude mcp add-json mcp-debugger …
Client setup →

languages

Seven languages, one contract

Every adapter drives the language's own production debugger through the Debug Adapter Protocol — the same engines your IDE would use, without the IDE.

LanguageEngineLaunchAttachRemote attachNotable
Pythondebugpylogpoints, exception breakpoints
JavaScript / TypeScriptjs-debugVS Code's engine, TS auto-detection
Rubyrdbgattach into containers & pods
RustCodeLLDBCargo-aware launch, logpoints
GoDelvenative DAP, logpoints
JavaJDI bridgeJDWP attach, class hot-swap
.NET / C#netcoredbgPID attach, portable PDB handling

Plus a mock adapter for testing agent integrations without any toolchain installed.

choosing a debug server

mcp-debugger vs. an IDE-bound debug server

Microsoft's DebugMCP exposes VS Code's debugger over MCP and is a good choice when your agent works inside a running VS Code. The projects make different structural trade-offs:

mcp-debuggermicrosoft/DebugMCP
Runs headless (CI, containers, k8s, cloud)✓ standalone processrequires running VS Code
Transportsstdio + Streamable HTTPHTTP (localhost)
Distributionnpx, npm, DockerVS Code Marketplace
Remote attach without an IDE✓ debugpy / rdbg / JDWP
Per-session process isolationshares the VS Code instance
Java hot-swap
In-IDE debugging UX beside the agentplanned✓ native
C/C++, PHP✓ via VS Code extensions

If your agent runs in a terminal, a pipeline, or a cloud sandbox — or needs to reach a process on another machine — you want mcp-debugger.

agent skill

Tools say what an agent can do. The skill teaches it to debug well.

mcp-debugger ships an agent skill: when to reach for the debugger, the session golden path, root-cause bisection discipline, and the per-language quirks we learned so your agent doesn't have to. The server also serves condensed guidance in-band via MCP instructions and a debugging-workflow prompt.

$ npx skills add debugmcp/mcp-debugger

who stands behind this

Built in the open. Accountable on paper.

mcp-debugger is stewarded by Sycamore LLC and led by John Franklin. AI agents write most of the code; a human maintainer makes every merge, release, and security decision. The project is MIT-licensed — the grant doesn't depend on the steward.

Governance

Supply chain

  • Sigstore provenance on every npm package — verify with npm audit signatures @debugmcp/mcp-debugger
  • SPDX + CycloneDX SBOMs attached to each GitHub release
  • OIDC trusted publishing, SHA-pinned CI actions — full controls
  • OpenSSF Scorecard · Best Practices