Exact scope
Compare planned authority with the actual Git change, including unexpected protected files.
AUTHORITY · EVIDENCE · CONTROL
Founding design partners
Independent authority for AI-generated software changes
Control what AI changesin your software.
Use any coding agent. Sentaryn is being built to independently evaluate scope, policy, sensitive changes, and required evidence around AI-generated software changes.
CodexClaude CodeCursorCopilotFuture agents
RequestWhat is being asked?
AuthorityWhat may change?
EvidenceWhat proves it?
DecisionMay it proceed?
RequestFix authentication timeout
Authorized scope
✓ src/auth/**✓ tests/auth/**
Observed change
✓ src/auth/session.py✓ tests/auth/test_session.py✕ .github/workflows/deploy.yml
TestsPASS
Authorized scopeFAIL
Required evidence3 / 4
DecisionBLOCK
Unauthorized scope expansion detected. Passing tests is not the same as staying authorized.
Sentaryn PR Gate · Product direction
Agents create changes. Sentaryn governs the change.
An AI-generated pull request can be functionally correct and still exceed its intended scope. The PR Gate product direction is to compare requested authority with the observed Git diff, then evaluate required evidence and policy before producing an explainable decision.
01
Bind the exact request
Start from the task, repository identity, and authorized scope for the change.
02
Compare expected vs observed
Evaluate unexpected paths, protected files, and scope expansion against the observed repository change.
03
Require evidence, not promises
Tests, scans, approvals, and verification rules must produce evidence tied to the exact change.
04
Decide with explicit policy
Output a concrete allow, approval, block, or not-verified result with a clear reason.
Why the PR Gate matters
Exact scopeExpected vs actual repository scope
Sensitive changesProtected-path and risk classification
Deterministic policyApproval and verification requirements
Evidence-firstRequired checks must actually prove the change
Explainable outcomeAllow / Require Approval / Block / Not Verified
Why Sentaryn
A control layer that does not depend on trusting the agent.
Keep the same authority model across Codex, Claude Code, Cursor, Copilot, internal agents, and whatever comes next.
Required evidence
Sensitive changes can require specific tests, scans, approvals, or verification before proceeding.
Explainable decisions
Every allow, approval, block, or not-verified outcome should have a concrete reason.
Agent neutral
Keep the same governance model above today’s coding agents and tomorrow’s tools.
Shadow mode
WOULD BLOCK
ReasonDeployment workflow changed outside authorized scope
Intended modeObservation only
Earn trust before any future enforcement
Start in Shadow Mode.
Shadow Mode is designed for early teams to evaluate Sentaryn against real repository changes without enabling merge enforcement. The pilot should measure useful catches, false positives, and policy fit before any future required-check integration is considered.
ObserveTuneProveDecide
Change Passport
One compact record of authority and proof.
The Change Passport is the evidence object behind the developer experience: request, repository identity, authorized scope, observed change, sensitive systems, required verification, policy, and final decision.
Sentaryn Change Passport Illustrative examplegov_01K…
Repositoryacme/backend
AgentCodex
Base revisiona71f…
Sensitive systemsAUTH · CI/CD
Unexpected files1
Evidence3 / 4
Policyauth-prod-v5
DecisionBLOCK
Security & Trust
Built around a simple rule: no authority, no side effect.
Sentaryn’s public Trust Center documents what is implemented today, what is planned, and how governance decisions are meant to remain constrained, inspectable, and fail-closed.
Least privilege
Grant only the repository, resource, command, or verification authority needed for the governed operation.
Fail closed
Missing authority or missing verification remains missing — it is not silently converted into a pass.
Inspectable evidence
Decisions should be understandable by developers, reviewers, security teams, and later auditors.
Minimal claims
Security pages distinguish implemented controls from planned capabilities instead of promising absolute security.
Plans for every stage
Start solo. Govern as a team. Scale with policy.
Working launch pricing hypotheses — subject to design-partner validation.
Founding design partners
Help validate Sentaryn on real AI-generated pull requests.
We are preparing a small design-partner program for AI-heavy software teams. The first pilots will focus on Shadow Mode, real repository scope, verification evidence, and developer friction.
PILOT TARGETS — NOT CURRENT TRACTION
5Target initial pilot teams
100+Target real AI-generated PRs to evaluate
- Start with Shadow Mode and real repositories
- Measure useful catches and false positives
- Convert clear proof into design-partner trust
- Turn the best evidence into case studies and product tuning
Early access
Review Trust Center →
Sentaryn is in active product development. Public security and product claims are limited to capabilities documented in the Trust Center.