OpenClaw 2026.10.10 gives operators more visibility into active work and adds new ways to review or talk to an agent. The November 3 stable release puts progress in Chat Details, introduces experimental Advisor reviews and adds a split speech option for browser Talk. Official release notes
Where did chat progress go?
The Details panel brings together session information and task activity. It can open alongside a wide conversation and can be toggled manually when space is limited. Subagent activity includes running and finished workers, with transcripts available for inspection. Chat Details changes
Start a harmless task in a fresh session after updating. Open Details while the task runs, then narrow the window and reopen the panel. If you routinely delegate work, check that you can identify which worker produced a result before acting on it. For a stalled run, compare the activity, transcript and actual destination state before retrying.
What can experimental Advisor do?
Advisor can review ongoing work and suggest a correction when the agent appears to be drifting or repeating an unsuccessful approach. It is enabled from Settings → System → Labs and uses additional model calls. Its suggestions can be mistaken. Tool actions continue to follow the configured approval rules. Advisor release notes
Test it on a task with a clear finish line. Record the initial instruction, the point at which advice appears and whether the advice improves the outcome. Include a run that is proceeding correctly so you can see whether the reviews introduce unnecessary interruptions. Check the cost and elapsed time alongside correctness.
What is split speech in browser Talk?
Split speech routes a spoken request to the current chat agent and uses separately configured services for transcription and speech generation. It requires microphone access in a secure browser context. Realtime voice switching and camera input are outside this mode. Split speech release notes
Check voice in stages. Start with a short spoken question and confirm the transcription, then check the agent’s reply in chat and listen to the playback. This separates microphone or transcription trouble from an agent error or a speech output problem. Use non-sensitive test audio while verifying which providers process each part.
Check the workflows you rely on
Follow the update path for your installation and verify a usable backup before changing the running service. The related 10.9 release covers recovery and security changes for operators moving forward from an older version. Updating documentation · 10.9 release notes
After the update, confirm the Gateway is running the expected version, send a normal message and check one representative workflow. If you use scheduled or event-triggered actions, run a controlled test with a destination you can inspect. Keep a record of the input and the resulting action so you can distinguish successful delivery from a progress indicator.
For recovery checks, see the earlier OpenClaw security and recovery guide. FindMilan provides OpenClaw setup and upgrade support, with a related agent setup project and workflow automation services.
Sources
Release facts come from the official notes. The verification steps are FindMilan’s suggested acceptance checks.
Frequently asked questions
What changed in OpenClaw 2026.10.10?
The release adds Chat Details, experimental Advisor reviews and an optional split speech path for browser Talk.
Can Advisor approve an action?
Advisor can suggest course corrections. The operator and configured approval rules retain control of tool actions.
What does split speech need?
Configure transcription and speech generation and grant microphone access in a secure browser context.
What should I check after upgrading?
Confirm the running Gateway version, inspect progress in a test session and exercise the voice or automation paths you actually use.