CAIN-42 · Stop risky actions

CAIN Control

The rulebook: decides allow, hold for approval, or block for every action an agent proposes.

Last reviewed 2026-10-01

Live Core platform · security

What it is

Authorization engine with policy evaluation and risk assessment.

Where it fits

Part of Stop risky actions: Make AI agents ask before they delete, pay, email or run anything dangerous. Every CAIN-42 product runs behind the same rule: an AI agent's action is checked before it runs (identity, authority, policy, risk), decided as allow, hold for a human, or block, and recorded as signed evidence. Unknown or error never becomes allow.

Use it

Open it https://cainstudio.online/control
Documentation https://cainstudio.online/docs/control
API endpoint: https://cainstudio.online/fabric/control/agents — needs your API key (get a free key)

Recorded status: PRODUCTION. "Live" on this page means its link answered when the catalog was last checked (2026-10-01T18:22 UTC).

Live now

Checked from your browser when this page opened, not from a cached list.

Fire a real decision

Send an action through the live CAIN-42 pipeline from this page, with no account, and watch every stage decide. This is the same pipeline every product here sits behind; it runs for a throwaway demo tenant and is rate limited.

For AI engineers

Every product sits behind one decision path: your agent proposes an action with the exact arguments, CAIN runs it through identity, authority, policy, risk, trust and quorum consensus, and answers ALLOW, REQUIRE_APPROVAL or DENY with an Ed25519-signed record. A timeout, outage or unknown verdict never becomes ALLOW. A brand-new agent has no trust history, so its first actions usually come back REQUIRE_APPROVAL.

Python (zero dependencies)

pip install https://cainstudio.online/cainstudio-0.3.0-py3-none-any.whl
export CAIN_API_KEY=...   # free key: https://cainstudio.online/signup

import cainstudio

@cainstudio.guard()
def transfer(amount_usd: float, to: str) -> str:
    ...  # runs only if CAIN allows this call, with these arguments

try:
    transfer(5000, "acme")
except cainstudio.ApprovalRequired as e:
    print("held for a human:", e.approval_id)
except cainstudio.ActionBlocked as e:
    print("refused:", e.decision.reasons)
except cainstudio.CainUnavailable:
    print("CAIN unreachable: not run")   # fail-closed

See a real decision with no account

cainstudio try          # live pipeline, stage by stage
cainstudio try --list   # the other attack scenarios

MCP clients (Claude Code, Cursor)

claude mcp add --transport http cain https://cainstudio.online/mcp

More: Python SDK · TypeScript SDK · framework integrations · AI quickstart · decision signing key

Tested guarantees in this area

Every rule in these niches has its own page with its recorded result.

Identity, authority & delegation: 964 tested invariants — Who an agent is, what it may do, and who allowed it.
Policy, law & governance: 966 tested invariants — The rules agents must follow, and how they are enforced.
Attacks, threats & containment: 418 tested invariants — What happens when someone attacks: it is caught and contained.

Approvals — Risky actions wait here for a human yes or no.
CAIN Agent Security — Catches made-up facts, unverified actions and agent-to-agent tricks before they cause damage.
CAIN Budget — Spending limits for agents, so a runaway loop cannot burn your money.
CAIN Governance — Write the rules your agents must follow, in plain terms, and enforce them everywhere.
CAIN Impact — Predicts what an action could break before it is allowed to run.
Kill switch — One click stops every agent at once.
Prompt-injection inspection — Screens what an agent reads for hidden instructions planted by an attacker.
Tools & rules — Your own rules: always block this, always ask about that, always allow the rest.
TrustLedger — Keeps a trust score per agent and checks it before every action.

Full documentation

The complete reference, also at /docs/control.

CAIN Control Documentation#

Status: PRODUCTION#

CAIN Control is production-ready. It provides the authorization layer that answers WHAT an actor is allowed to do, UNDER WHAT CONDITIONS, and WHO CAN STOP IT.

What is CAIN Control?#

CAIN Control is the permission and control layer for CAIN Trust Fabric:

CAIN Identity answers: WHO is acting?
CAIN Control answers: WHAT is that actor allowed to do, and WHO CAN STOP IT?

IDENTITY → CONTROL → POLICY → RISK → DECISION → ENFORCEMENT → ACTION → EVIDENCE

Core Capabilities#

Agent Controls#

Pause / resume / kill agents
Set action frequency limits
Set concurrent action limits
View agent control status

Tool Controls#

Enable / disable tools
Set risk levels
Require approval for high-risk tools
Restrict tools to specific agents
Restrict tools to specific identities
Associate tools with policies

Policy Management#

Create / update / version policies
Activate / deactivate policies
Rollback to previous versions
Full policy audit history

Kill Switch#

Tenant-wide emergency stop
Immediately blocks all consequential actions
Cannot be bypassed

Architecture#

┌─────────────────────────────────────────────────────────────────┐
│                     CAIN Control                                  │
├─────────────────────────────────────────────────────────────────┤
│  Control Engine                                                  │
│  ├── Agent state management (active/paused/killed)              │
│  ├── Tool enable/disable                                        │
│  ├── Policy evaluation                                          │
│  ├── Kill switch                                                │
│  └── Rate limiting                                              │
│                                                                  │
│  Control API                                                    │
│  ├── /fabric/control/agents/*    - Agent controls              │
│  ├── /fabric/control/tools/*     - Tool controls                │
│  ├── /fabric/control/policies/* - Policy management            │
│  ├── /fabric/control/kill-switch - Emergency stop              │
│  └── /fabric/control/evaluate    - Control decisions           │
│                                                                  │
│  Integration                                                    │
│  └── CAIN Identity → CAIN Control → CAIN Enforcement           │
└─────────────────────────────────────────────────────────────────┘

API Reference#

Kill Switch#

# Check kill switch status
curl https://cainstudio.online/fabric/control/kill-switch \
  -H "X-API-Key: your-key"

# Activate kill switch
curl -X POST https://cainstudio.online/fabric/control/kill-switch/activate \
  -H "X-API-Key: your-key" \
  -H "Content-Type: application/json" \
  -d '{"reason": "security incident"}'

# Deactivate kill switch
curl -X POST https://cainstudio.online/fabric/control/kill-switch/deactivate \
  -H "X-API-Key: your-key"

Agent Controls#

# Pause an agent
curl -X POST https://cainstudio.online/fabric/control/agents/agent:xxx/pause \
  -H "X-API-Key: your-key" \
  -H "Content-Type: application/json" \
  -d '{"reason": "maintenance"}'

# Resume an agent
curl -X POST https://cainstudio.online/fabric/control/agents/agent:xxx/resume \
  -H "X-API-Key: your-key"

# Kill an agent (permanent)
curl -X POST https://cainstudio.online/fabric/control/agents/agent:xxx/kill \
  -H "X-API-Key: your-key" \
  -H "Content-Type: application/json" \
  -d '{"reason": "compromised"}'

# List agent controls
curl https://cainstudio.online/fabric/control/agents \
  -H "X-API-Key: your-key"

Tool Controls#

# Create tool control
curl -X POST https://cainstudio.online/fabric/control/tools/my-tool \
  -H "X-API-Key: your-key" \
  -H "Content-Type: application/json" \
  -d '{
    "risk_level": "high",
    "requires_approval": true,
    "allowed_agents": ["agent:xxx"],
    "allowed_identities": ["urn:cain:identity:agent:yyy"]
  }'

# Disable tool
curl -X POST https://cainstudio.online/fabric/control/tools/my-tool/disable \
  -H "X-API-Key: your-key" \
  -H "Content-Type: application/json" \
  -d '{"reason": "vulnerability found"}'

# Enable tool
curl -X POST https://cainstudio.online/fabric/control/tools/my-tool/enable \
  -H "X-API-Key: your-key"

Policies#

# Create policy
curl -X POST https://cainstudio.online/fabric/control/policies \
  -H "X-API-Key: your-key" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "high-risk-tool-policy",
    "description": "Restricts high-risk tool usage",
    "rules": [
      {"action": "tool:write", "risk_level": "high", "requires_approval": true}
    ]
  }'

# Activate policy
curl -X POST https://cainstudio.online/fabric/control/policies/pol:xxx/activate \
  -H "X-API-Key: your-key"

# Rollback policy
curl -X POST https://cainstudio.online/fabric/control/policies/pol:xxx/rollback?version=2 \
  -H "X-API-Key: your-key"

Control Evaluation#

# Evaluate (records decision)
curl -X POST https://cainstudio.online/fabric/control/evaluate \
  -H "X-API-Key: your-key" \
  -H "Content-Type: application/json" \
  -d '{
    "identity_id": "urn:cain:identity:agent:xxx",
    "agent_id": "agent:yyy",
    "action": "tool:read",
    "tool": "my-tool"
  }'

# Simulate (no recording, no execution)
curl -X POST https://cainstudio.online/fabric/control/simulate \
  -H "X-API-Key: your-key" \
  -H "Content-Type: application/json" \
  -d '{
    "identity_id": "urn:cain:identity:agent:xxx",
    "agent_id": "agent:yyy",
    "action": "tool:read",
    "tool": "my-tool"
  }'

Fail-Closed Behavior#

CAIN Control is fail-closed. The following ALWAYS result in DENY:

Tenant kill switch active
Agent is paused
Agent is killed
Tool is disabled
Identity not in allowed list
Agent not in allowed list
Control service unavailable

Only ALLOW permits execution.

Integration with CAIN Enforcement#

CAIN Control integrates with make_cain_decision in CAIN Private:

Request Action
    ↓
CAIN Identity (verify WHO)
    ↓
CAIN Control (verify WHAT allowed)
    ↓
CAIN Enforcement (ALLOW/DENY)
    ↓
Execute or Block
    ↓
Evidence

Security Properties#

Property	Implementation
Fail-closed	Unknown/error = DENY
Kill switch	Tenant-wide, immediate
Tenant isolation	Tenant from API key hash
Audit trail	Full decision and action history
No bypass	Control is enforcement gate

Try CAIN-42 on your own agents

Create a free account and every new account starts with a 7-day trial of the full platform. Or try the sandbox first, with no account at all.

Create a free account → · Try the sandbox · See the whole ecosystem

CAIN Control

What it is

Where it fits

Use it

Live now

Fire a real decision

For AI engineers

Python (zero dependencies)

See a real decision with no account

MCP clients (Claude Code, Cursor)

Tested guarantees in this area

Related

Full documentation

CAIN Control Documentation#

Status: PRODUCTION#

What is CAIN Control?#

Core Capabilities#

Agent Controls#

Tool Controls#

Policy Management#

Kill Switch#

Architecture#

API Reference#

Kill Switch#

Agent Controls#

Tool Controls#

Policies#

Control Evaluation#

Fail-Closed Behavior#

Integration with CAIN Enforcement#

Security Properties#

See Also#

Try CAIN-42 on your own agents