Control every action
your AI agents take.
Cryptographic identity, runtime policy, scoped credentials, and tamper-proof audit — in a single sidecar. Built for regulated finance.
Auto-advances every 5s · click any badge or list item to jump · ← → toggle phase
From agent intent to provable action.
Four checkpoints. ~12ms median. Fail-closed by default.
Identity
Sidecar resolves the agent's cryptographic identity. KMS-signed, DID-bound, on-chain.
agent_id: "did:regent:agent:0x4a3f8b..." bound_to: "neobank.com/loan-officer-bot" attestation: SPIFFE x509 SVID
Policy
Cedar policy evaluated against tool, operation, table, amount, context.
permit when {
resource.table in ["customers", "loans"]
&& context.amount_usd < 50000
}Access
Scoped JWT minted for exactly one downstream call. Seconds-long expiry.
scope: "neobank:loans:write" exp: now + 60s jti: "auth_9f2a1b..."
Audit
Event hashed, Merkle-batched, anchored on Solana. Provable to regulators.
event_id: "evt_a3f9b2..." merkle_root: "sha256:..." solana_tx: "5wKpZ..."
Watch a NeoBank loan officer agent hit the gate.
Real policy, real mandate enforcement, real revocation. Click through each agent action.
Click "Run next action" to start the decision flow.
Three agents. One control plane.
Same Regent Control deployment. Different policies. Different mandates.
Loan-officer copilot
Agent reads customer history, checks credit, drafts approvals. Regent Control gates PII tables, enforces $50k/transaction cap, escalates anything above to a human manager via CIBA.
- Cedar policy: per-table access by role
- On-chain mandate: per-tx + daily + monthly caps
- CIBA escalation to manager phone
Algorithmic trading agent
Agent executes BTC/USDT trades on Binance based on signal models. Mandate enforces position size, Guardian flags anomalous behavior, revocation kill-switch works in <1s.
- Mandate enforcement before exchange API
- Guardian ML risk score per trade
- Instant revoke on Solana anchor
Tier-1 support agent
Agent answers tickets, looks up policy details, issues refunds under $200. Regent Control caps refund authority, blocks PII export, full audit trail per ticket for compliance.
- Refund mandate with $200 ceiling
- PII column-level deny
- Per-ticket Merkle-anchored audit
Pay for the agents you ship.
Audit storage scales with usage. On-chain anchoring rolled in.
Developer
For prototypes, internal tools, and learning the platform.
Start free- SDK access (TS + Python)
- 1 agent + 1 mandate
- 100k audit events / month
- Community support
Team
For teams shipping agents to production with light governance.
Start trial- Everything in Developer
- 25 agents + unlimited mandates
- 10M audit events / month
- Docker sidecar (Phase 2 beta)
- MCP gateway access
- Email support, business hours
Enterprise
Regulated finance, healthcare, government. SAML/OIDC/SCIM, dedicated CSM, SOC 2 evidence package.
Contact sales- Everything in Team
- Unlimited agents
- Unlimited audit events
- Envoy filter (Phase 3)
- SAML 2.0 / OIDC / SCIM
- Dedicated mTLS workload identity
- On-prem deployment option
- 24/7 support + SLA
All tiers include the on-chain Solana audit anchor at no additional cost. Storage overage at $0.10 / 1M audit events.
Get a Regent Control deployment guide and a 30-min architecture review.
For regulated-finance teams shipping AI agents in 2026. We'll respond within 48 hours.
By submitting, you agree to receive product updates from Regent Protocol. Unsubscribe anytime.