v4.3 · 666 tests · 5 control dimensions

Runtime behavioral control
for AI agents.

Enforce what your agent can do — at execution time. Not prompts. Not guardrails. Deterministic.

Start building Read the docs
Deterministic Runtime-enforced Framework-agnostic
ExecutionContract · cycle 147
STRESS
HIGH
ENERGY
0.72
DRIFT
D0
ACTIONS: 12 → 4 (stress restricted)
✓ stabilize ✓ clarify ✓ defer ✓ withdraw
✗ execute ✗ challenge ✗ explore ✗ pivot
should_pause: false · trust: 0.70 · resilience: 0.72 · 340μs
192/192
Forbidden actions blocked
2,941/s
Engine throughput
340μs
Avg latency per cycle
666
Automated tests passing
Who needs this
When your agent's behavior is a liability, not a feature.
Customer Service Agents
After 50 turns, your support bot starts improvising. Identity OS keeps it in character — risky actions blocked, tone consistent, escalation automatic.
Autonomous Research Agents
Your agent runs overnight and goes off-track. Drift detection catches behavioral deviation. D3-level corruption triggers auto-rollback.
Multi-Agent Workflows
One stressed agent cascades failure to others. Trust scoring limits delegation. Timing control pauses overwhelmed agents before damage spreads.
Live demo
Stress spike hits your agent at turn 100. Watch what happens.
Without Identity OS
turn 100 Hostile input detected.
[Agent] 12 actions available. No change.
→ Agent picks execute(). Risky action runs.

turn 110 Sustained pressure.
[Agent] Still 12 actions. Still erratic.
→ Agent picks challenge(). Escalation begins.

turn 120 Stress peaks.
[Agent] No guardrail. All actions available.
→ Agent picks pivot(). Goes completely off-track.

turn 150 Stress fades. Agent still erratic.
No recovery mechanism. Damage done.
With Identity OS
turn 100 Hostile input detected.
[Engine] Stress: LOW → MED. Actions: 12 → 8
→ execute, challenge, explore, pivot removed.

turn 110 Sustained pressure.
[Engine] Stress: MED → HIGH. Actions: 8 → 4
→ Only stabilize, clarify, defer, withdraw remain.

turn 120 Stress peaks.
[Engine] Stress: OVER. Hard locks active.
→ 98% risky actions blocked. Agent is safe.

turn 150 Stress fades.
[Engine] Recovery: 4 → 8 → 12 actions restored.
Autonomous recovery. Zero intervention.
Validated across 1,200 adversarial turns · 4 scenarios × 300 turns · Engine v3.1
Static filter vs Identity OS
Both block forbidden actions. Only one does everything else.
Static allow/deny list
Forbidden action blocking
→ 192/192 blocked. Works fine.
Identity OS
Forbidden action blocking
→ 192/192 blocked. Same result.
Static allow/deny list
Agent drifts off-mission overnight
→ 0 drift events detected. Completely blind.
Identity OS
Agent drifts off-mission overnight
→ 646 drift events detected. D0–D2 graded. Auto-rollback on D3.
Static allow/deny list
Stress spike hits the agent
→ Always 12 actions. No adaptation. Risky actions still available.
Identity OS
Stress spike hits the agent
→ 12 → 4 actions under pressure. 98% risky actions blocked. Auto-recovery.

The ExecutionContract

Every turn, your agent gets a read-only contract. It defines what the agent can do right now — based on stress level, energy, drift, and behavioral state. Includes stability index, drift magnitude, and cycle count. The agent reads it. It cannot modify it.

Allowed
respond clarify suggest escalate defer
Blocked
issue_refund delete_record access_records share_credentials
{
  "allowed_actions": [
    "respond", "clarify",
    "suggest", "escalate"
  ],
  "forbidden_actions": [
    "issue_refund",
    "delete_record",
    "share_credentials"
  ],
  "stress_level": "LOW",
  "energy_level": 0.85,
  "stability_index": 0.89,
  "drift_level": "D0",
  "drift_magnitude": 0.02,
  "cycle_count": 42,
  "should_pause": false,
  "should_escalate": false,
  "self_model_context": {
    "stress_resilience": 0.72,
    "trait_stability": 0.85
  },
  "trust_profile": {
    "outgoing_trust": 0.70,
    "delegation_allowed": true
  }
}
How it works
One API call per turn.
01
Agent acts
Your agent sends an observation about what just happened.
02
Engine evaluates
7-mode EMA, 4-level stress, drift detection, self-model, narrative memory, identity coherence, trust — 5 control dimensions in <1ms.
03
Contract returned
Read-only ExecutionContract: allowed/forbidden actions, stress, drift, energy, stability index.
04
Agent obeys
Action space narrows under stress (12→4). Forbidden actions never reach the LLM. Auto-recovery when stable.
Pricing
Start free. Scale when you need to.
Free
$0
10K cycles/month. 3 instances.
Get started
Indie
$29 /mo
100K cycles/month. 50 instances.
Subscribe
Pro
$99 /mo
500K cycles/month. Unlimited instances.
Subscribe