Unified events and cases
Events and Cases provide a shared entry point for monitoring signals, inventory context, and the diagnostic process.
Open source / v0.2.0 released
AgenticOps
AgenticOps is an open-source event governance and diagnostic workbench for network operations. It brings NetBox, ELK, Zabbix and other data sources into a workflow that connects Events, Cases, multi-agent analysis, operational memory and controlled execution.
Who it is for
- Network and infrastructure teams unifying NetBox, ELK, and Zabbix signals
- AIOps practitioners who need durable evidence, state, and human approval boundaries
- Developers evaluating multi-agent workflows in real operations environments
What problem it solves
Alerts, logs, and asset context are spread across systems, while diagnosis often remains in temporary conversations and human memory instead of a recoverable, reviewable evidence trail.
Core capabilities
Persistent multi-agent graph
The asynchronous multi-agent graph persists runtime state so diagnostic work can be recovered and traced.
Evidence, hypotheses, and timeline
Evidence, Hypothesis, and Timeline records connect conclusions to their sources, reasoning, and sequence.
Searchable operations memory
Operations memory preserves prior diagnostic context for later retrieval and reuse.
NetBox, ELK, and Zabbix integrations
Integration boundaries for NetBox, ELK, and Zabbix correlate inventory, logs, and alerts.
Human Gate and observe-only controls
High-risk actions are not executed automatically; Human Gate, observe-only, and Shadow Mode define the validation boundary.
Typical workflows
- Receive or import a monitoring event
- Group related events into a tracked Case
- Use the multi-agent graph to gather evidence, form hypotheses, and update the timeline
- Add operations memory and inventory context
- Plan or run controlled actions behind a Human Gate
Use cases
- Triage incidents across monitoring and inventory systems
- Organize root-cause hypotheses and timelines around evidence
- Correlate NetBox, ELK, and Zabbix context
- Plan controlled remediation behind human approval boundaries
Inputs
- Zabbix alerts and monitoring events
- ELK logs and search results
- NetBox inventory, IP, and topology context
- Operator-provided Case context
Outputs
- Unified Event and Case records
- Evidence, root-cause hypotheses, and diagnostic timeline
- Recoverable agent graph state
- Remediation plans and audit records behind human approval
Deployment and data boundary
- v0.2.0 runs through Docker Compose in the user's own server environment
- Production access should use an external HTTPS reverse proxy and independent secrets in `.env`
- NetBox, ELK, Zabbix, and model connections are user-configured and data remains in the target environment
- Start with observe-only acceptance before Shadow Mode or controlled execution
Clone and check out the release
git clone https://github.com/Jaycelu/agenticops.git
cd agenticops
git fetch --tags --prune
git checkout v0.2.0
git describe --tags --exact-match
git rev-parse HEADPrepare environment variables
cp deploy/docker.env.example .env
openssl rand -hex 32 # write to APP_SECRET_KEY
openssl rand -hex 24 # write to POSTGRES_PASSWORD
# Also set AUTH_PUBLIC_BASE_URL and FRONTEND_URL
# Keep AUTOMATION_OBSERVE_ONLY=True
# Keep AUTH_COOKIE_SECURE=True and WEBHOOK_ALLOW_HTTP=False
chmod 600 .envValidate and build images
docker compose config -q
if docker compose config | grep -E 'change_me|replace_with'; then
echo 'Unresolved configuration found; stop deployment'
exit 1
fi
docker compose buildStart services and run migrations
docker compose up -d postgres
docker compose run --rm migrate
docker compose up -d backend worker frontend
docker compose psCreate the first break-glass administrator
export BOOTSTRAP_ADMIN_PASSWORD='a-long-random-password-from-your-password-manager'
docker compose exec -e BOOTSTRAP_ADMIN_PASSWORD="$BOOTSTRAP_ADMIN_PASSWORD" backend \
python -m scripts.bootstrap_admin \
--username admin \
--display-name Administrator \
--confirm-create-first-admin
unset BOOTSTRAP_ADMIN_PASSWORDVerify health endpoints
curl -fsS http://127.0.0.1:8000/health/live
curl -fsS http://127.0.0.1:8000/health/ready
curl -fsS http://127.0.0.1:8000/health/dependencies
curl -fsS http://127.0.0.1:8000/metrics
docker compose exec worker python -m scripts.check_worker_healthSystem requirements
- Linux server with Docker Engine 24+ and Docker Compose v2
- Git, curl, and OpenSSL
- External HTTPS reverse proxy for production use
- Reachable NetBox, ELK, Zabbix, and model services as enabled
Current version and update
0.2.0
Published：2026-07-16
Content updated：2026-07-16
Known limitations
- v0.2.0 does not automatically execute high-risk actions
- The frontend currently uses polling rather than SSE or WebSocket
- Interactive Human Gate resume remains in the follow-up roadmap
- Real NetBox, ELK, and Zabbix integrations require validation in the target environment
Frequently asked questions
Does AgenticOps automatically change production devices?
No high-risk action runs automatically by default. v0.2.0 uses Human Gate, observe-only, and Shadow Mode to define validation and execution boundaries.
Must NetBox, ELK, and Zabbix all be deployed?
No. Enable the data sources available in your environment, then validate connectivity and data quality for each real integration.
Does the website provide an AgenticOps installer?
The v0.2.0 Release contains source code and no binary assets, so the website links the formal tag, deployment guide, and Docker Compose startup steps without inventing an installer.
Related technical guides
No complete English guide is published for this product yet.
View all English guides · Browse the related Chinese knowledge base
Product screenshots
