OpenClaw just shipped v2026.3.8, and the headline is simple: it’s getting easier to operate agents safely in real workflows.
1) New CLI backup + verify commands (practical ops win)
The release adds openclaw backup create and openclaw backup verify to help you archive and validate your local OpenClaw state. The feature supports config-only backups (via --only-config), optional workspace inclusion, and manifest/payload validation so you can detect corruption before you need a restore. This is a small change with big “production hygiene” impact—especially if you iterate fast or run multiple agents across machines.
2) Provenance metadata + “receipt” injection for ACP runs
Another notable improvement is optional provenance handling for ACP ingress. With openclaw acp --provenance off|meta|meta+receipt, agents can retain and report where ACP-origin context came from, including session trace IDs. In practice, this makes it easier to debug multi-hop agent runs and to explain why an agent made a decision—useful for teams that need auditability.
3) Web search tooling quality-of-life improvements
v2026.3.8 also tightens the web search experience. Provider ordering is now alphabetized across selection and onboarding, and there’s an opt-in Brave mode (tools.web.search.brave.mode: "llm-context") that returns extracted grounding snippets with source metadata—handy when you want citations and traceability in research-heavy agent workflows.
How to apply this today
- Before an upgrade: run a config-only backup so you can roll back cleanly.
- After major config edits: use verification to catch broken archives or drift early.
- For compliance-sensitive use cases: enable provenance receipts for clearer audit trails.
If you’re building repeatable operations around OpenClaw, these changes are worth adopting. For more OpenClaw resources, see our OpenClaw News hub.
Source: OpenClaw v2026.3.8 release notes.


