cd /news/ai-safety/show-hn-fengarde-ocsf-native-open-so… · home › topics › ai-safety › article
[ARTICLE · art-159988] src=supermhel.github.io ↗ pub= topic=ai-safety verified=true sentiment=↑ positive

Show HN: Fengarde–OCSF-native open-source SIEM with local AI triage (Apache-2.0)

Developer supermhel released Fengarde, an Apache-2.0 open-source SIEM that ingests raw log lines and produces verified incidents plus draft NIS2 notifications, with local AI triage built in. Fengarde runs as nine services communicating only over a message bus, ships 46 detection rules tagged to MITRE ATT&CK, ATLAS or ICS, and includes a demo in which an SSH brute force and a DNS burst sent over unauthenticated UDP port 5514 become a real alarm; the repository's SSOT.md separates proven capabilities from stubbed or unbuilt ones.

read2 min views2 publishedOct 29, 2026
Show HN: Fengarde–OCSF-native open-source SIEM with local AI triage (Apache-2.0)
Image: source

Evidence out.

FENGARDE turns a raw log line into a verified incident and a draft NIS2 notification.

Departures #

pipeline · lab stack Idle: nothing replayed

Departure board: pipeline stages and the alarms raised by the replayed attack
Time Source Event Track Status
--- --- --- --- ---

Track= bus topic.

One pipeline. Receipts included. #

Seven things FENGARDE does that a log viewer does not. Each names the place in the repository where you can check it.

Nine services. One bus. #

No service calls another. Each reads topics from the bus and writes topics back, so any stage can be restarted, swapped or scaled alone. Every box opens its directory on GitHub, and during a replay the boxes light as the attack passes through them.

Swipe sideways to see the whole bus.

What runs. What does not. #

Every capability gets the board's honest word. Anything stubbed, opt-in or not built says so.

Capability board
Service What Interface Status
--- --- --- ---

46 rules, all listed. #

Read straight from the rule files in the repository: level, title and the MITRE ATT&CK (or ATLAS, or ICS) tag each one carries.

Twelve views. Open any. #

Each tile opens that view of the real console on the recorded data.

One take, no cuts. #

Recorded in a scripted browser against a running stack. A real SSH brute force and a DNS burst go in over UDP 5514 at about 0:30.

Run it in ten minutes. #

$ git clone https://github.com/supermhel/fengarde.git
$ cd fengarde && make preflight
$ make demo          # an SSH brute force becomes a real alarm

$ cd tools/console_e2e && npm install
$ node check.mjs       # every view as admin, analyst, read-only
$ node tour.mjs out    # re-records the video, sends the attack
  • Docker Desktop with enough memory for OpenSearch. make preflight checks ports and limits before the first start.
  • The console, its checks and this page's video all come from the repository: services/ws7-dashboard/app ,tools/console_e2e .
  • The UDP syslog listener is unauthenticated: expose port 5514 on a trusted network or VPN only.
  • Status, proven versus claimed: SSOT.md . Features per view:console feature matrix .
── more in #ai-safety 4 stories · sorted by recency
── more on @fengarde 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/show-hn-fengarde-ocs…] indexed:0 read:2min 2026-10-29 · —