ProtocolProbe Self-Test MCP server
Commit Nov 3, 2026Runs automated MCP server security and protocol checks against your stdio-based MCP server within a GitHub Actions CI workflow.
Also known as Sultan57git/protocolprobe-selftest-action
- Stars
- 0
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 3, 2026
- License
- ISC
What is ProtocolProbe Self-Test MCP?
ProtocolProbe Self-Test runs automated MCP server security and protocol checks against your stdio-based MCP server within a GitHub Actions CI workflow.
Summary generated from the project's README
About ProtocolProbe Self-Test MCP server
Automating security and protocol validation for your MCP server within a GitHub Actions CI workflow is what this tool enables. It works by starting your stdio-based server directly from your repository and running checks inside your own runner. This ensures that no data is sent to external parties, as all results remain within your workflow logs or optional JSON and SARIF reports.
Developers can configure the action using a specific command to launch their server, such as a node command or a uv command for Python servers. The process includes options to enable fuzzing with malformed inputs or to allow mutating tool calls on disposable test servers. By using the sarif-anchor input, you can map findings back to your source code even when testing compiled build outputs. This setup provides a way to catch vulnerabilities or protocol errors during the development lifecycle without exposing your server to the public internet.
What you can do with it
- 1Detect security vulnerabilities in CIIdentify security findings and fail the GitHub Action step if a vulnerability is discovered.
- 2Integrate security findings with GitHubUpload security scan results directly to GitHub code scanning using SARIF files.
- 3Test server resilience to malformed inputEnable fuzzing to see how your tools handle unexpected or malformed data.
Written from the project's README. Every use case is checked against a quote from it.
Automated checks
Ran Nov 16, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachablesultan57git/protocolprobe-selftest-action is public and reachablegithub.com/sultan57git/protocolprobe-selftest-action ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)ISCspdx.org/licenses/ISC.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-03github.com/sultan57git/protocolprobe-selftest-action ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubProtocolProbe self-test (GitHub Action)
Runs protocolprobe-selftest against your own MCP server
in your CI: the workflow starts your server over stdio from your repository, the checks run inside your runner, and
the results stay in your workflow (log, optional JSON report, optional SARIF for GitHub code scanning). Nothing is sent
to ProtocolProbe.
Use it only on a server you own or are authorised to test. The action passes --i-own-this-server: adding it to your
workflow is that statement.
Usage
name: MCP self-test
on: [push, pull_request]
permissions:
contents: read
security-events: write # only needed for the SARIF upload step
jobs:
selftest:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
- uses: actions/setup-node@949feb2413d6458794dcd2491c4babbbce0c15c1 # v7.1.0
with:
node-version: 22
- run: npm ci && npm run build
- uses: Sultan57git/protocolprobe-selftest-action@v1
with:
command: node ./dist/server.js
sarif-anchor: src/server.ts
- uses: github/codeql-action/upload-sarif@9b3254085494b55a0f7c4c2b3d2d52cc712ce370 # v4.39.1
# upload also when the self-test step failed because of a finding, but only if a SARIF file was written
if: always() && hashFiles('protocolprobe-selftest.sarif') != ''
with:
sarif_file: protocolprobe-selftest.sarif
category: protocolprobe-selftest
Python server (needs uv on the runner):
- uses: Sultan57git/protocolprobe-selftest-action@v1
with:
command: uv run --directory . server.py
Inputs
| input | default | meaning |
|---|---|---|
command |
(required) | Command line that starts your MCP server over stdio. Split on spaces; quoting is not supported, so put a complex command in a script file. |
version |
0.5.1 |
Exact protocolprobe-selftest version from npm, 0.4.0 or later (earlier versions lack --fail-on-finding and --sarif). Ranges and tags are refused. |
fail-on-finding |
true |
Fail the step (exit 5) when a check needs attention (VULNERABLE or FOUND). |
sarif |
protocolprobe-selftest.sarif |
SARIF 2.1.0 file to write. Empty to skip. |
sarif-anchor |
File the SARIF results point at (line 1). Default: the first file of command inside the working directory. Set it when that file is build output (for example dist/) that is not in the repository. |
|
out |
JSON report to write. Empty to skip. | |
allow-mutating |
false |
Also run the check that calls tools which may change data. Only on a disposable test server. |
allow-fuzzing |
false |
Also run the check that sends malformed input to your tools. |
working-directory |
. |
Folder to run in; paths in the other inputs are relative to it. |
Outputs: exit-code (0 result printed, 1 could not start or connect, 2 refused arguments, 5 finding with
fail-on-finding) and sarif-file.
What it does and does not do
- Starts the command you give, connects as an MCP client and runs the checks described in the package README. Checks that call tools which may change data, or that send malformed input, are off unless you turn them on.
- Stdio servers only. The package's
--urlmode needs a one-time proof added to the server's instructions; it is not offered here. - The score counts only checks that ran; a check that did not run says nothing about that behaviour. A clean result is not a certificate of safety.
- Network: the only outbound traffic is the npm download of the package, plus whatever your own server does. Your server is started with a minimal environment, so the runner's secrets are not passed to it.
- Inputs reach the script only as environment variables and are validated; the command is never run through a shell
string or
eval.
Versioning
v1 moves to the newest v1.x.y release; v1.x.y tags never move. To pin exactly, use the commit SHA:
uses: Sultan57git/protocolprobe-selftest-action@<full commit SHA> # v1.0.0.
Security
Report a problem with this action to hello@protocolprobe.com (see SECURITY.md).
Licence
ISC. ProtocolProbe: https://protocolprobe.com
ProtocolProbe Self-Test MCP questions
Can I test Python-based MCP servers?
Yes, you can test Python servers if you have uv installed on the runner.
What kind of server types are supported?
The action only supports stdio servers.
Does the action run tools that might change my data?
By default, checks that call tools which may change data are turned off unless you explicitly enable them using the allow-mutating input.