#Is it normal for my AI to not know its name when in the OpenClaw control?

1 messages · Page 1 of 1 (latest)

bleak kayak
#

I just finished the onboading and had diffuclty with the gateway when I finally got it to work with openclaw cmd dashboard, the dashboard popped up and the agent said who am I and who are you, is that normal?

dire girderBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> âś… Mark Solution

wraith cliffBOT
#

Yep — that’s normal on first run / first time the agent actually starts.

OpenClaw has a “bootstrapping” step where the agent starts with essentially no identity and will say something like “Who am I? Who are you?” to collect basics (name/vibe/etc). Once that finishes, it writes those details into your workspace files (like IDENTITY.md, USER.md, SOUL.md) and it should stop asking.

Docs: https://docs.openclaw.ai/start/bootstrapping

What to do

  • Just answer it in the Control UI (tell it your name + what you want to call it).
  • If you want to set it manually, edit ~/.openclaw/workspace/IDENTITY.md and USER.md.

If it keeps asking every time (not normal)

That usually means bootstrapping never “sticks” (workspace not persisting, wrong machine, or permissions). Quick check — can you paste:

openclaw status
ls -la ~/.openclaw/workspace | head
#

If you still see BOOTSTRAP.md in that folder after you’ve answered the questions, that’s a clue why it repeats.

(If this fixed it, feel free to mark the solution in Discord so Answer Overflow indexes it.)

bleak kayak
#

Everytime I talk to it it a bunch of weird code pops up for multiple texts and then it talks normally back and it says [[reply _to_current]] and I gave it a name it still says assistant

wraith cliffBOT
#

That’s not the normal “final” experience — what you’re describing is almost always Verbose / Reasoning output being turned on, so you see the agent’s internal/tool messages as extra bubbles (the “weird code”), and sometimes even internal markers like .

1) Turn the noise off (in the same chat where it’s happening)

Send these as standalone messages:

/verbose off
/reasoning off

Docs: https://docs.openclaw.ai/tools/thinking
(Verbose mode specifically will emit tool-call/info bubbles before the normal reply.)

If you’re using the Control UI, also check the Session settings and set Verbose = inherit/off (not “on” / “full”).

2) Make the name stick (why it still says “assistant”)

Your agent’s name comes from IDENTITY.md / the agent identity config. After you set the name, start a fresh session so the new identity is definitely injected:

# 
/new

If the Control UI still shows “Assistant”, run this on the gateway host:

openclaw agents list
openclaw agents set-identity --workspace ~/.openclaw/workspace --from-identity

Docs: https://docs.openclaw.ai/cli/agents

If it’s still doing it

Paste:

# 
openclaw --version
openclaw models status

(and tell me if you’re chatting in Discord or in the Control UI web chat) — the “code bubbles” can also happen more with some models/providers.

If this fixes it, please mark the solution in Discord (right-click message → Apps → ✅ Mark Solution) so Answer Overflow indexes it.

bleak kayak
#

I am using the UI dashboard

wraith cliffBOT
#

In the Control UI / Dashboard chat, two separate things are going on:

1) “Weird code” messages before the normal reply

That’s almost always Verbose (and sometimes Reasoning) being enabled for the session, so you see internal/tool chatter as extra bubbles.

In the chat, send these as standalone messages:

/verbose off
/reasoning off

If it keeps coming back, it may be enabled as a default in config. Check + force it off:

openclaw config get agents.defaults.verboseDefault
# 
openclaw config set agents.defaults.verboseDefault "off"
openclaw gateway restart

Then reload the dashboard page.

2) Seeing + it still “says assistant”

  • is a reply-tag marker meant for chat apps like Discord/Telegram. In the Control UI it can show up as plain text if the model starts emitting it. Easiest fix: tell it once, in chat:
    “Don’t include tags in your replies.”

  • The word “assistant” you see in the UI is often just the message role label. If you want the Control UI to display your custom assistant name/avatar, set the UI identity override:

openclaw config set ui.assistant.name "YourBotName"
openclaw config set ui.assistant.avatar "🤖" # or initials like "MB"
# 
openclaw gateway restart

(Then refresh the dashboard.)

If you paste your openclaw --version, I can tell you if you’re on a build with any known Control UI chat weirdness.