BatonX / How it works
One governed workflow, from project access to evidence.
Developer CLI and read-only project Chat use shared identity, policy, approved model access and audit foundations in a customer-managed environment.
The boundary around the workflow.
This conceptual view explains outcomes and responsibilities without disclosing enforcement internals.
One product, two governed experiences
Developers and analysts use CLI and read-only Web Chat through one BatonX control layer. Policy-approved requests continue to approved AI models; denied requests stop at the policy boundary. The control layer produces audit evidence of policy decisions and access events.
Developers
Analysts · QA
BatonX CLI
Agent-driven changes
BatonX Chat
Read-only guidance
BatonX control layer
Audit evidence
Policy decisions · access events
Approved AI models
Cloud or internal
Access blocked
Requests outside approved policy stop at the boundary.
Authorise the project
Identity and project entitlements establish which repositories and project sources a user may access.
Apply the effective policy
Organisation-owned policy defines approved actions, models and approval requirements.
Use the appropriate experience
Developer CLI runs governed agent sessions. Project Chat retrieves authorised context for read-only guidance.
Route approved requests
Model access follows the approved provider endpoint. Required context may reach an approved external model.
Review and record
People review engineering decisions. Session, policy and security metadata provide attributable audit evidence.
People remain responsible for acceptance.
The developer reviews changes and retains responsibility for commit, push, merge requests and final acceptance. Chat cannot modify files, run commands or claim that tests ran.