Overview

Live
Data Sources
-
configured
Sources Up
-
connected
Services
-
discovered
Services Healthy
-
live
MCP Endpoint: http://localhost:3000/mcp

How it works

AgentMCP call
Gateentitlement · RBAC · catalog
ConnectorsPrometheus · Loki
Analysisscored health

Sources

Loading...

Services

Loading...

Needs attention

Loading…

Services by health score

Loading…

Data Sources

All sources

Loading...

Services

Discovered services

Loading...

Connectors

Installed connectors

Loading...

Available from the Connector Hub

Open Hub ↗
Loading...

Upload a connector bundle

Install a signed connector .tgz directly — handy for air-gapped environments. The bundle is always verified against the configured trust root before it is loaded.

Service Health

Loading health data...

Infrastructure Topology

Sources & counts

Loading topology...

Grouped by host (pivots on the RUNS_ON relation)

Scope = any resource pointed to by an IN_NAMESPACE edge (e.g. k8s namespaces, future: vCenter folders).
Loading...

Layered graph

click a resource to inspect · drag to reposition · wheel to zoom · drag the background to pan
Tab to focus · Enter to inspect · arrows to move focus · Esc to clear

Inspect — agent activity

Whether a call is normal vs. the learned baseline for that identity — behavioral guardrails. Distinct from Policies (allowed at all).

Flows identities → tools → backends · edge width = call volume · colour = allowed / deviation / blocked

Tab to focus · Enter to inspect · arrows to move focus · Esc to clear
Loading flows…
Mode

Behavior profile learn a baseline from observed traffic, then accept the rules that look right

Review queue

No suggestions yet. Click Learn from traffic to derive rules from what agents have actually done.

Accepted rules

No accepted rules yet. Accept suggestions above to build the enforceable profile.

Rejected rules

No rejected rules. Rejected suggestions are kept here (not deleted) and won't reappear in the review queue on re-learn — restore one anytime.

Deviations calls outside the accepted profile — would-block in dry-run, blocked in enforce

Switch to Dry-run and accept some rules — calls that fall outside them show up here as “would-block”, so you can verify the profile before enforcing.

Settings

How often the agent scans for anomalies

Configure how service health scores are calculated. Weights must sum to 1.0.

Weights

Thresholds

CPU (%)

Error Rate (req/s)

Latency P99 (seconds)

Log Errors (/min)

Status Boundaries

Each data source has its own metric definitions with backend-specific queries (PromQL, LogQL, etc.). Use {{service}} as placeholder for the service/job name.

NameUnitQueryDescription
Select a source above

Access Control

Who can sign in and which role they hold — identities, SSO/OIDC, sessions. Authentication.

How governance fits together

Access Controlwho you are Productswhat you're offered Policieswhat you're allowed Inspectis it normal
Each layer answers a different question. Catalog runs alongside, adding resource context (owners, criticality). Audit Log records every change. Policies = allowed at all (static); Inspect = normal vs. the learned baseline (behavioral).

Roles & bindings

Loading…
Use Edit policy above — no command line needed. The block below is the optional API equivalent for automation/CI.

Postmortems

Generated reports

Loading…

Tool Playground

Invoke a tool

Context Products

Curated, governed tool bundles you expose to an agent or credential.

Which tools a credential is offered (packaging). Distinct from Policies, which decide what's permitted.

MCP Products

Loading…
Bind a credential to a product

Bind a credential to a product via OMCP_KEY_PRODUCTS; the agent's next /mcp session then sees only that product's tools:

OMCP_API_KEYS="agent:tok_ops,ci:tok_dev"
OMCP_KEY_PRODUCTS="agent=ops-bundle;ci=dev-bundle"

Full docs →

Legacy catalog (enterprise)

Deprecated. Backed by OMCP_ENTERPRISE_CATALOG_FILE. Use MCP Products above for new deployments.

Loading…

Policies

What a role is allowed to call at all — coarse RBAC (resource × action). Distinct from Inspect, which judges whether a call is normal.

Probe a permission

Roles

Loading…
Select a role on the left to see its permission matrix.

Audit Log

An append-only, tamper-evident record of every mutating action — who did what, when.

Recent decisions

Loading…

Management changes

Loading…

Entitlement

Read-only

Access-control gate

Loading…