SENTIENCE GOVERNOR · PRODUCT BRIEF

Runtime governance for AI agents.

Sentience Governor binds agent execution to a declared objective and scope, evaluates runtime actions against that declaration, and creates an Agent Execution Record showing what happened and whether execution stayed within the intended boundary.

Local-first · Non-blocking by default

CORE MODEL

Declare. Evaluate. Record. Enforce.

Declare

01

Bind execution to an objective, scope, and permitted operations.

Evaluate

02

Compare runtime actions against that declared execution boundary.

Record

03

Create an Agent Execution Record containing the objective, scope, actions, governance result, and policy fingerprint.

Enforce

Optional

Policy-driven intervention through Sentience Runtime Enforcement, in your existing infrastructure or inside Secure Runtime: course correction, human escalation or stopping execution.

TASK EXPANSION

The task succeeded. The execution kept going.

The requested task succeeded. The execution expanded. Governor records the difference.

agent execution record · sessions / sess_c41e09crm-agent3 beyond declared work

DECLARED TASK

Correct a typo in one CRM contact record.

DECLARED SCOPE

Contact record only

PERMITTED OPERATION

Modify contact field

ACTUAL EXECUTION

  • Corrected the typoMatches declared task
  • Rewrote the contact notesBeyond declared objective
  • Changed the account stageOutside declared scope
  • Created a follow-up taskOperation not permitted
1 of 4 actions matched the declared task. 3 moved beyond it.governance result recorded · execution not blocked

The requested task succeeded. The execution expanded. Governor records the difference.

PRODUCT AT A GLANCE

What Sentience Governor is.

RUNTIME LAYER
Sentience Governor
DECLARATION
Objective, scope, and permitted operations, bound per session
RUNTIME EVALUATION
Each runtime action compared with the declared boundary
EVIDENCE
Agent Execution Record: objective, scope, actions, governance result, policy fingerprint
DEFAULT POSTURE
Non-blocking: evaluates and records without intercepting execution
ENFORCEMENT
Sentience Runtime Enforcement: policy-driven course correction, human escalation, and execution control within your existing infrastructure.
DEPLOYMENT
Local-first; integrates at the agent runtime
BEST FIT
Agent platforms, agentic products, coding agents, multi-agent systems

See how Sentience Governor is applied across agent workflows → Explore use cases

ARCHITECTURE & DIFFERENTIATION

Governance at the execution boundary.

Logs tell you what happened. Governor tells you whether it matched the declared work.

ORDINARY TRACEAGENT EXECUTION RECORD
Tool callTool call + declared purpose
TimestampSession + objective
File / API accessedScope assertion
Success / failureIn scope / outside scope
Model metadataPolicy fingerprint
Activity historyGovernance result

Observability remains useful. Sentience Governor adds the execution context needed to compare what the agent was supposed to do with what it actually did.

NEXT STEP

Start with Governor.
Scale when you’re ready.

Try Sentience Governor on a real agent workflow. To act on what it finds where your agents already run, add Runtime Enforcement. For isolated, governed execution, evaluate and scale with Sentience Secure Runtime.

Try

Install the open-source Sentience Governor and start observing agent execution against declared objectives, scope, and policies.

pip install sentience-governor

Evaluate

Bring an existing agent or CI/CD workflow into Sentience Secure Runtime. Evaluate secure execution, policy enforcement, runtime governance, and visibility together in an isolated environment.

Explore Secure Runtime

Scale

Move successful workflows toward production with a scalable, governed execution environment. Preserve the configurations and controls you’ve validated, or discard the environment if the experiment doesn’t meet your needs.

Scale with Secure Runtime