#Lix for AI Agents

AI agents generate a tremendous volume of changes. These changes need to be reviewed, attributed, and approved; Lix change control provides the safety layer so every edit is traceable, humans decide what ships, and any change can be rolled back.

#Every change is reviewable

• Use attribution to see exactly which agent (or human) changed a field, paragraph, or table cell.
• Compare versions with entity-aware diffs so reviewers understand the intent of a change instead of guessing from raw text.
• Keep a durable audit trail; you can query history for “Which agent touched this configuration last?” at any time.

#Users stay in control

Agents can draft changes, but humans stay in the loop with lightweight review tools.

• Change proposals bundle the diff, discussion, and approval state. Accept the good ideas, reject the hallucinations.
• Start review conversations with stakeholders or other agents using comments and mentions from conversations.
• Merge once the proposal is ready, or request another revision—the entire workflow maps to familiar pull-request style collaboration.

#Safe sandboxes for agents

Versions let you spin up isolated environments so agents can explore ideas without touching production data.

• Run multiple agents in parallel, each in their own version.
• Compare their diffs, merge the best outcome, or discard a version entirely.
• If something slips through, use restore to jump back to a known-good state in seconds.

#Typical workflow

1. Orchestration layer opens a fresh version for an agent task.
2. Agent writes changes; Lix auto-commits and tracks attribution.
3. Another agent or a human reviewer inspects the diff and opens a change proposal.
4. Reviewers comment, request edits, and approve when ready.
5. Merge the proposal or discard it; restore any previous snapshot if plans change.

#Coming soon: automated guardrails

#Next steps

• Walk through the Getting Started guide to wire Lix into your agent pipeline.
• Learn how to diff, merge, and experiment with versions.
• See change proposals in action in the live example.