Read the system as a graph
Agents can traverse from a symptom to related logs, metrics, traces, alerts, configs, hosts, services, workloads, incidents, and ownership context.
// cirdan.layer
Cirdan is the AI-native observability layer built on LogForge Unicron.
Cirdan is an emerging premium layer for teams building AI-operated infrastructure. It turns logs, metrics, traces, events, alerts, configs, and live state into connected operational knowledge that agents can inspect, cite, explain, report on, and use for approved action.
graph knowledge
agent-oriented operations
human-approved action
built on Unicron
// what.cirdan.is
An observability and action layer for agents, not another dashboard.
Cirdan is designed for a world where AI agents become primary readers of infrastructure state. Instead of leaving the model to reconstruct operations from isolated screens, Cirdan keeps a continuously updated system graph that reflects what is running, what changed, what failed, and what evidence supports each answer.
Turn signals into grounded answers
Operational data is compressed into connected knowledge, so reports and explanations can cite the state and relationships that shaped the result.
Keep action scoped and governed
Cirdan can stay advisory, or move into action only where a human operator has explicitly allowed a scoped change, remediation, or setup task.
// why.agents.need.knowledge
Dashboards show state. Agents need operational memory.
Human operators can bridge gaps between panels, runbooks, deployment history, alerts, and tribal knowledge. Agents need that context represented directly, with relationships intact.
DASHBOARD-FIRST
views
- Logs, metrics, alerts, and config live in separate places.
- Context is rebuilt manually during every incident.
- The model sees snippets, not operational relationships.
- Action depends on a human translating diagnosis into commands or rule changes.
CIRDAN
graph
- Signals are connected to services, workloads, hosts, incidents, and live state.
- Historical and current context stays available for the next agent task.
- The model can inspect, cite, explain, report, and recommend from one operational knowledge layer.
- Approved action can be routed through scoped policy instead of improvised terminal work.
// agent-native.observability
Built for AI agents, not human operators.
Traditional observability was built for human operators: dashboards, charts, log streams, and manual context-building. Cirdan keeps that evidence available, then reshapes it into context an agent can read, cite, and reason across while systems are live.
// human.dashboard
fragments
Human-oriented observability
Dashboards, charts, alerts, and logs expose slices of state. The operator still has to stitch them into incident context.
charts
alerts
api p95 high
queue lag rising
logs
checkout timeout
retry budget hit
configs
rule: page after 5m
deploy: payments:v42
dashboards
alerts
logs
configs
human eyes
agent perception
// cirdan.context
agent-ready
Cirdan
Cirdan turns observability into agent infrastructure: a living context layer agents can inspect, connect, cite, and use to prepare approved action.
connected.signals
service + workload + alert
citations
logs, deploy, rule state
action.prep
approval required
inspect
connect
cite
prepare
// context.pipeline
operator governed
- inspect read current state and evidence
- connect preserve service relationships
- cite keep answers grounded in sources
- prepare stage scoped action for approval
// operating.modes
Advisory by default. Action only when explicitly allowed.
Cirdan separates analysis from authority. It can explain what is happening without changing the system, then step into bounded action only after human approval and policy allow it.
Explain, report, investigate
Use Cirdan to summarize incidents, compare service behavior, identify likely causes, explain container or workload context, and produce operational knowledge for teams and end users.
Scoped work under operator control
When a human explicitly permits it, Cirdan can help set up monitoring, change rules, run approved remediation, and take scoped or full actions on the operator's behalf.
// built.on.unicron
Cirdan sits above the Unicron control plane.
LogForge Unicron is the Docker-native base: local and agent-forwarded logs, metrics, Docker events, alert rules, notifications, file access, and safe remediation. Cirdan builds on that foundation as a premium AI-native layer for graph knowledge and agent-oriented operations.