The Clyro Capabilities Map.

Loop detection, cost ceiling, step limit and business-logic guardrails. On by default for every wrapped agent.

PolicyEvaluator, AuditLogger, LoopDetector and CostTracker for every MCP tool call. Default-deny enforcement under the hood.

Author rules in YAML or the dashboard. Push policy from cloud to every machine the agent runs on.

Suggests policy rules based on patterns Clyro sees in your agent's actual runs. One-click promote from suggestion to enforcement, right from the dashboard.

Govern Bash, Edit and Write tool calls inside Claude Code.

An append-only trace of every blocked or allowed action with the rule, the input and the outcome.

Step through any agent run after the fact. Inspect inputs, decisions, tool calls and outputs in order.

Append-only JSONL with sensitive-field redaction. Logging never blocks execution (fail-open).

Stream traces to your existing OTel collector. Plays with Datadog, Honeycomb, Grafana Tempo and the rest.

Alerts when an agent's behaviour moves away from its baseline. Catches the silent regression nobody noticed.

A shared view of every agent, every trace, every policy across your team.

Admin, member and viewer roles. Org → team → project → agent hierarchy on the Enterprise tier.

See which concerns and regulations are covered by policy, which are gapped, and which are firing in production.

Notifications fire where your team already works.

A 5-dimension reliability score per agent, derived from policy coverage, violations, drift, replay fidelity and cost stability.

Track Agent Reliability Index over weeks and months. See whether a change improved or degraded the agent.

Export of policy + evidence + ARI bundled for a third-party reviewer. Targeting Enterprise tier.

Audit in progress. We will publish the report when it lands.

Run Clyro inside your own VPC. Available on the Enterprise tier when shipped.