who's in charge · payments → Sarah Kim (AP Manager) · data → Marcus Webb (Platform Lead) · email → Priya Patel (Support Lead) · escalation → David Chen (CFO) / Ana Duarte (CISO) · demo roles: production authority syncs from your identity provider
⚡ Needs a human
Approve releases it · Reject… blocks it and the agent re-plans off your reason · Take over hands the task to you · Ask why makes the agent justify itself · Escalate sends it up the org chart
Activity
Everything your agents tried, decided and finished, plus the hash-chained audit trail underneath.
History
Audit trail
Corridors
A grade per department, from your own ledger. Owners come from your org chart; scorecards are per corridor.
Corridor cards fill in as actions flow: grades update live from your ledger. In guided demo mode, run a few scenarios from Overview first.
Three steps. Under ten minutes. Week one runs in shadow mode and costs nothing.
1 · Wrap each MCP server
In your client's config, replace the server command with:
node gateway.mjs -- <your original command>
Works for Claude Code, Codex, Kimi, LangChain: anything that speaks MCP. Keys go in ripcord.vault.json: agents never see them.
2 · Choose where asks land
Slack approvals with buttons: see SLACK-SETUP.md (10 minutes). Or start with this dashboard: every ask lands in Needs you either way, routed to the corridor's owner from ripcord.config.json.
3 · Run week one in shadow
RIPCORD_SHADOW=1 observes without gating anything. Then read your Near-Miss Report: what it would have held, and what those mistakes would have cost.