Developer guide
Recipes and CI
Ready-made ways to run MCPatch from a terminal, a script or a scheduled job.
#A shell script
Prints the verdict and each step that needs attention, links the report, and exits non-zero when the connector fails. It needs curl and jq.
#!/usr/bin/env bash
set -euo pipefail
url="${1:?usage: mcpatch-check.sh https://host/mcp}"
api="https://mcpatch.monster13liar.xyz"
tmp="$(mktemp)"
code=$(curl -sS -o "$tmp" -w '%{http_code}' -X POST "$api/api/check" \
-H 'content-type: application/json' \
-d "$(jq -n --arg u "$url" '{url: $u}')")
if [ "$code" != "200" ]; then
echo "MCPatch answered $code: $(cat "$tmp")" >&2
exit 2
fi
jq -r '.verdict.headline' "$tmp"
jq -r '.steps[] | select(.status == "fail" or .status == "warn")
| "\(.status | ascii_upcase) \(.title)\n \(.summary)\n fix: \(.fix)"' "$tmp"
echo "Report: $api/r/$(jq -r '.id' "$tmp")"
[ "$(jq -r '.verdict.status' "$tmp")" != "fail" ]Exit codes: 0 for a pass or a warning, 1 when a step fails, 2 when MCPatch itself could not be reached or refused the request (for example the free limit).
#Node.js
const url = process.argv[2]
const res = await fetch('https://mcpatch.monster13liar.xyz/api/check', {
method: 'POST',
headers: { 'content-type': 'application/json' },
body: JSON.stringify({ url }),
})
if (!res.ok) throw new Error(`MCPatch answered ${res.status}: ${await res.text()}`)
const r = await res.json()
console.log(r.verdict.headline)
for (const s of r.steps.filter((x) => x.status === 'fail' || x.status === 'warn')) {
console.log(`${s.status.toUpperCase()} ${s.title}\n ${s.summary}\n fix: ${s.fix}`)
}
console.log(`Report: https://mcpatch.monster13liar.xyz/r/${r.id}`)
process.exit(r.verdict.status === 'fail' ? 1 : 0)#A scheduled GitHub Actions job
Runs the script once a day and on demand. The job fails when the connector fails, so GitHub emails you.
name: MCP connector check
on:
workflow_dispatch:
schedule:
- cron: '0 7 * * *'
jobs:
check:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- run: bash scripts/mcpatch-check.sh https://mcp.example.com/mcp#After a deploy
Run the script as the last step of your deploy pipeline, after the new version is live. A failing step tells you the deploy broke the connection before your users find out. Because a deploy can change your tools, the same run also reports tool changessince the previous check.
#Servers behind a token
Add token (and headerName if it is not Authorization) to the request body. Keep the token in a secret of your CI system, never in the repository. It is sent over HTTPS, used for that one run and not stored.
jq -n --arg u "$URL" --arg t "$MCP_TOKEN" '{url: $u, token: $t}' | curl -sS -X POST "$api/api/check" \
-H 'content-type: application/json' -d @-