See what your AI agents actually do at runtime
Sentience Governor compares what an agent was asked to do with the actions it actually takes, so you can see when the work moves outside the task.
$ pipx install sentience-governorNo account. No API key. No telemetry by default. Your governance traces stay on your machine.
AI agent governance and execution tracing for modern applications
Why Runtime Governance
See When an Agent Goes Beyond the Work You Gave It
Governor connects the task and scope the agent starts with to the actions it takes at runtime. When the work moves outside that context, you get a clear record of what happened.
Guardrails
Shape or evaluate model output.
Runtime Governance
Compare agent actions with the objective and scope declared for that session.
Observability
Record system and application activity.
Declare the task → Evaluate each action → Record the evidence
What Happens When an Agent Acts Without Clearly Declaring Its Task
- Declared
- 3,840 tokens
- Undeclared
- 1,000 tokens
- Flag
- INTENT_MISSING · POL-001
Get Started
From Install to Governed Traces in Minutes
Start with the agent framework you already use. No hosted account is required.
Install Governor → Attach it to your runtime → Run the agent → Inspect the record
Works with
Claude Code
Available nowExecution hooks and slash commands.
MCP
Available nowClient wrapper; the optional server requires explicit opt-in.
LangChain
Available nowCallback handler integration.
LangGraph
ExperimentalExperimental middleware integration.
Trust
Local by default. Narrow by design. Auditable in practice.
Local by default
Run Governor without an account or external service. Sentience sends no telemetry by default.
Focused on declared context
Governor evaluates the intent, scope, and context the agent declares for the task. Governor evaluates raw payload content only when an integration explicitly provides it.
Auditable output
Governor records governance events in sequence so teams can inspect what happened during the run. The runtime is open source under Apache 2.0.
From One Agent to a Governed Fleet
Start Locally, Standardize as a Team, and Add Central Control
Developer
Available nowInstall Governor locally and record structured evidence from a real agent run.
Team
Available nowStandardize governance profiles and review execution records across the team.
Enterprise
DirectionBring shared policies, fleet visibility, and enforcement across your agent environment. Gateway integration remains optional.
Bring Us an Agent Workflow You Want to Govern
Show us the agent, what it is supposed to do, and where you need more visibility or control.
Talk to us →·Explore pricing →Technical Publication
Latest from Governor’s Log
Aug 29, 2026
What’s Your Claude Up To?
We pointed Sentience Governor at our own Claude Code history.
Aug 26, 2026
When Nothing Obviously Breaks
The logs were valid. The reports looked right. The meaning was wrong.
Aug 25, 2026
When a Simple Instruction Becomes an Expensive System
The biggest risk with capable AI agents may not be disobedience. It may be competent execution built on the wrong framing.