Is it suitable for a trial?
CF-Server-Monitor suits a controlled trial when you want host monitoring on Cloudflare and can verify the deployment yourself. Our September 28, 2026, run at commit dfb9bf1 installed successfully, but its build and test commands failed. That leaves a local execution problem to investigate before enrolling a fleet; it does not establish that a Cloudflare deployment would fail.
What runs on Cloudflare
The English README describes an Agent on each server reporting metrics to a Worker. The Worker stores data in D1 and pushes updates through Durable Objects and WebSocket. Documented capabilities include CPU, memory, disk and network metrics, history charts, and offline, resource and expiration alerts. These are project descriptions, not deployed features verified by our run.
The README says the Agent provides no WebSSH, remote command delivery, or controller channel. That boundary may suit operators who do not want remote administration through their monitoring dashboard. It is not a security guarantee established by our testing. The documented Docker command installs the Agent, rather than a local container stack for the Cloudflare backend.
What the stored run establishes
Installation succeeded in 15 seconds. The build command failed after 4 seconds, and the test command failed after 7 seconds. Both returned exit code 1 without timing out.
Recorded parser summaries report 48 passing tests, 8 failing tests, and 56 total tests. The supplied footer reports zero skipped tests, while the aggregate skippedTests field is null. The counts have not been reconciled against complete logs and do not establish exhaustive unique coverage. The available tail does not show the individual failure messages.
The build log gives a concrete error: ReferenceError: File is not defined in undici while Vite loads its configuration. It reports Node.js v18.20.8. Installation warnings show newer engine requirements, including Node >=22.0.0 for Wrangler and >=22.19.0 for one recorded undici version. Check the runtime against the installed dependencies before repeating the build. No supplied rerun verifies that changing Node resolves the build or test failures.
The sandbox image is labeled lab-node:22, but the execution log reports Node.js v18.20.8. Preserve that distinction when investigating the environment. The build banner also says cf-server-monitor@2.8.5, while the README supplied for the same commit lists changes under 2.8.6. This review identifies the tested code by commit dfb9bf1 rather than assigning it a reconciled release version.
The run did not establish successful Cloudflare deployment, Agent reporting, browser behavior, authentication, alert delivery, retention behavior, or fleet capacity. The testing methodology explains the publisher's current measurement process; it does not prove that this older run used the current procedure.
The audit records disagree
The dependency summary records 102 packages and 309 MB. The raw install tail separately says it added 85 packages and audited 86. These figures describe different recorded outputs; the supplied evidence does not reconcile them.
The stored audit summary records zero advisories, but the install tail reports three high-severity vulnerabilities. Neither output establishes a clean security result. Scan coverage is unknown because scannedPackages is absent. Reconcile the audit output before relying on either count.
Deployment decisions before adding hosts
The README documents deployment through a GitHub fork connected to Cloudflare Workers or through GitHub Actions. The Actions route requires CF_API_TOKEN, CF_ACCOUNT_ID, D1_DATABASE_ID, and API_SECRET. After deployment, the admin panel generates installation commands for each host.
API_SECRET is initially both the Agent reporting secret and the admin password. The admin password can be separated later. Changing API_SECRET requires redeploying the Worker and updating Agents, so decide how to manage it before installing across the fleet.
The README estimates roughly 60 servers at a 60-second reporting interval and describes doubling capacity at 120 seconds as theoretical. Those are project estimates, not our capacity measurements. Check actual Cloudflare usage with your reporting interval, dashboard traffic, and history queries before depending on free-tier operation.
The Access issue has a later resolution report
In issue 109, the reporter confirmed on September 29 that the dev Beta9 fix resolved the refresh loop after Cloudflare Access returned to /admin. The issue was closed on September 30. This is a later upstream report, separate from our September 28 sandbox result. Check your deployed revision and authentication path; the report does not verify our tested commit or your deployment.
Choose alternatives by monitoring role
Beszel uses a PocketBase hub and an Agent on each monitored system. Its README documents Docker statistics, historical data, and alerts. Consider it as a host-monitoring replacement when you prefer that architecture or need its documented container metrics. No comparative benchmark or migration was tested here.
Uptime Kuma documents HTTP(s), TCP, WebSocket, Ping, DNS record, and push uptime checks. Consider it when endpoint availability is the main question, or alongside host monitoring to check services separately. This is a recommendation based on the documented monitoring roles, not a tested integration.
What the ratings assess
Setup ease and documentation use an editorial scale from 1 to 5. For setup, 1 means installation is blocked, 2 means installation succeeds but build or tests fail, 3 means local commands succeed without a verified deployment, 4 means deployment is verified with remaining manual work, and 5 means the documented setup path is verified through reporting and alerts. Setup ease is 2: installation succeeded, but build and test commands failed, and deployment remains unverified.
Documentation scores reflect written coverage: 1 covers only a project description; 2 adds basic installation; 3 includes deployment and configuration; 4 also covers Agent operation and maintenance; 5 adds security guidance and troubleshooting. Documentation is 5 because the supplied Chinese and English instructions cover those subjects. This score does not certify that the instructions execute correctly or resolve the runtime discrepancy.
Community and maturity are unassessed. Issue 109 shows a maintainer response and reporter confirmation for one defect, but that does not establish adoption or dependable support. The supplied sources do not independently establish the historical star count or release absence previously used in these ratings. The failed local commands remain relevant to setup, while deployment reliability and broader project maturity remain unverified.

