#status failed: Error: gateway not connected
supple pollen
trim heraldBOT
Short answer: no, there isn’t a separate TUI token cache/config.
openclaw tui reads the normal config via loadConfig().
Given your report, this is most likely one of these:
- CLI config vs service config mismatch (different profile/state dir)
- TUI is effectively in URL-override path (when
--urlis used, it won’t auto-pull token) - gateway.mode is
remote, so TUI looks atgateway.remote.token, notgateway.auth.token
Can you run these and paste output?
openclaw --version
openclaw gateway status --deep
openclaw config get gateway.mode
openclaw config get gateway.auth.mode
openclaw config get gateway.auth.token
If gateway status shows Config (cli) != Config (service), fix is usually:
openclaw gateway install --force
openclaw gateway restart
Also double-check you’re not launching TUI through an alias/wrapper that injects --url (that path requires explicit --token).
Docs:
If this is still reproducible on latest after the config-path check, it’s likely a bug worth filing: https://github.com/openclaw/openclaw/issues
supple pollen
https://pastebin.com/mzKgjNt3 neuer inhalt
trim heraldBOT
Nice report — this is super useful.
Quick answers to your 4 questions (based on current docs/runtime behavior):
- Protected config paths (for
config.patch)
- As far as documented/supported: no user-configurable protected-path list yet.
- Best supported mitigation right now is to deny control-plane tools for that agent/surface:
- deny
gateway(and usuallycron) via tool policy. - Relevant doc: https://docs.openclaw.ai/gateway/security (Control plane tools risk section)
- Batch-change guardrail
- There is currently no documented “N restart-required fields touched” guard.
- Existing safeguards are:
- control-plane write rate limit (3/60s per deviceId+IP),
- restart coalescing + cooldown.
- Doc: https://docs.openclaw.ai/gateway/configuration
- Restart deferral timeout
- Yes, this is configurable:
gateway.reload.deferralTimeoutMs - Default is 300000 (5 min).
- So you can raise it if you want less mid-run forcing.
MEMORY.mdtruncation
- This is configurable:
agents.defaults.bootstrapMaxChars(default 20000)agents.defaults.bootstrapTotalMaxChars(default 150000)agents.defaults.bootstrapPromptTruncationWarning(off|once|always)- Also recommended: keep
MEMORY.mdcompact and move long detail intomemory/*.md(on-demand retrieval). - Docs: https://docs.openclaw.ai/concepts/context and https://docs.openclaw.ai/reference/token-use
Given this includes feature requests + stability impact, please open a GitHub issue with that exact report text (it’s excellent as-is):
https://github.com/openclaw/openclaw/issues
If you want, I can help you turn it into a minimal repro + expected/actual format before you post.
supple pollen
Thanks it did work