Browser Control not working today
✅Solved
What I’m trying to do (goal + expected behavior)
I want my local OpenClaw agent to open/control Chrome or Brave (like it used to yesterday), then navigate/log into sites when instructed.
What happened instead (exact error text)
From my current session, the agent says it cannot open a browser and only has messaging/session tools available.
It reports that browser-control toolchain/relay access is not active in this runtime/surface.
What I ran / clicked
I asked the agent to open browser targets, e.g.:
“can you please open up a web browser on your end”
Agent checked available sessions and only found the current chat session (agent:main:main), no browser-controllable target/session.
This started while using OpenClaw through the current chat/control UI surface.
My environment (OS, install method, versions)
OS: Linux 6.8.0-100-generic (x64)
Node: v22.22.0
OpenClaw install path appears npm-global (~/.npm-global/lib/node_modules/openclaw)
Model/runtime shown in status: openai-codex/gpt-5.3-codex (main agent session)
Channel/surface at time of issue: webchat/telegram-linked control session
Relevant logs (smallest useful chunk)
Symptoms from agent:
“I can’t directly open a browser window from this chat session.”
“Right now, this session only has messaging/session tools, not browser-control toolchain.”
Session listing only returns current session, no controllable browser/relay session.
Extra context
This used to work previously (agent could open Chrome/Brave and continue tasks), so I’m trying to figure out what changed: capability gating, surface restrictions, relay not connected, or runtime/tool permission change.
I want my local OpenClaw agent to open/control Chrome or Brave (like it used to yesterday), then navigate/log into sites when instructed.
What happened instead (exact error text)
From my current session, the agent says it cannot open a browser and only has messaging/session tools available.
It reports that browser-control toolchain/relay access is not active in this runtime/surface.
What I ran / clicked
I asked the agent to open browser targets, e.g.:
“can you please open up a web browser on your end”
Agent checked available sessions and only found the current chat session (agent:main:main), no browser-controllable target/session.
This started while using OpenClaw through the current chat/control UI surface.
My environment (OS, install method, versions)
OS: Linux 6.8.0-100-generic (x64)
Node: v22.22.0
OpenClaw install path appears npm-global (~/.npm-global/lib/node_modules/openclaw)
Model/runtime shown in status: openai-codex/gpt-5.3-codex (main agent session)
Channel/surface at time of issue: webchat/telegram-linked control session
Relevant logs (smallest useful chunk)
Symptoms from agent:
“I can’t directly open a browser window from this chat session.”
“Right now, this session only has messaging/session tools, not browser-control toolchain.”
Session listing only returns current session, no controllable browser/relay session.
Extra context
This used to work previously (agent could open Chrome/Brave and continue tasks), so I’m trying to figure out what changed: capability gating, surface restrictions, relay not connected, or runtime/tool permission change.
Solution
To change it you edit the Gateway’s config (
Then start a new chat/session and try the browser command again.
Docs (tool profiles): https://docs.openclaw.ai/tools#tool-profiles-base-allowlist
openclaw.jsonQuick way (CLI)
On the machine running the Gateway:Then start a new chat/session and try the browser command again.
Docs (tool profiles): https://docs.openclaw.ai/tools#tool-profiles-base-allowlist
If it’s set per-agent (overrides global)
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
134,426Members
Resources
Was this page helpful?
