how to get my agent fill form online and book tickets for me? | Friends of the Crustacean 🦞🤝 | Page 1

kindred sun Apr 18, 2026, 7:28 PM

#

how to get my agent fill form online and book tickets for me?

scarlet fossilBOT Apr 18, 2026, 7:28 PM

#

viscid sailBOT Apr 18, 2026, 7:28 PM

#

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

scarlet fossilBOT Apr 18, 2026, 7:28 PM

#

cedar wharfBOT Apr 18, 2026, 7:30 PM

#

Yes, that’s the browser tool.

Quick path:

Check version first
pnpm openclaw --version
Start the agent browser
pnpm openclaw browser --browser-profile openclaw start
Open the site
pnpm openclaw browser --browser-profile openclaw open https://example.com
Log in yourself in that browser, then ask the agent:
“Use the browser to fill this form, stop before final submit/payment.”

Best practice:

#

Do not give the model your passwords
for sites where you’re already logged in, use the user browser profile
expect CAPTCHA, MFA, and payment/checkout to still need you

If browser seems missing, make sure:

browser.enabled: true
if you use plugins.allow, it must include "browser"

Docs:

If you want, paste your pnpm openclaw --version and I’ll give you the exact config for ticket-booking flows.

kindred sun Apr 18, 2026, 8:18 PM

#

i don't have any browser in my json file

cedar wharfBOT Apr 18, 2026, 8:18 PM

#

That’s normal, you usually won’t have a browser block unless you’re overriding defaults.

OpenClaw’s browser is documented as enabled by default, and the bundled browser plugin is also enabled by default.

If you want to make it explicit, add this to openclaw.json:

{
  browser: {
    enabled: true,
    defaultProfile: "openclaw",
  },
}

Important gotcha: if you use plugins.allow, it must include "browser" too:

#

{
  plugins: {
    allow: ["browser", "discord"],
  },
}

Then restart and test:

pnpm openclaw gateway restart
pnpm openclaw browser status
pnpm openclaw browser start

Docs:

#

If it still doesn’t work, paste:

pnpm openclaw --version
pnpm openclaw browser status

and I’ll help you narrow it down.

kindred sun Apr 18, 2026, 10:42 PM

#

"plugins": {
"entries": {
"openai": {
"enabled": true
},
"google": {
"enabled": true
},
"tavily": {
"enabled": true,
"config": {
"webSearch": {
"apiKey": "xxxxxx"
}
}
},
"browser": {
"enabled": true
}
}
},
"browser": {
"executablePath": "/usr/bin/brave-browser"
}

cedar wharfBOT Apr 18, 2026, 10:43 PM

#

Yep, that looks basically right for a managed Brave browser 👍

I’d make it a tiny bit more explicit:

browser: {
  enabled: true,
  executablePath: "/usr/bin/brave-browser",
  defaultProfile: "openclaw",
}

Then restart + test:

pnpm openclaw gateway restart

#

pnpm openclaw browser status
pnpm openclaw browser start
pnpm openclaw browser open https://example.com

If that works, the flow is:

open the site in the OpenClaw browser
log in yourself
tell the agent:
Use the browser to fill this form and stop before final submit/payment

A couple notes:

your current config is for the isolated OpenClaw-managed browser, not your already-open personal Brave session
that is actually the best place to start
for ticket buying, expect CAPTCHA / MFA / payment confirm to still need you

#

Docs:

If pnpm openclaw browser start errors, paste that output and pnpm openclaw --version and I’ll help from there.

kindred sun Apr 18, 2026, 10:49 PM

#

why browser.enabled = true, when plugins.browser.enabled = true is set... which one is correct.. i just want to avoid redundancy if it is there

cedar wharfBOT Apr 18, 2026, 10:50 PM

#

They’re different knobs.

plugins.entries.browser.enabled = load the bundled browser plugin
browser.enabled = turn the browser feature/runtime on

So the docs’ rule is basically:

plugin must be loaded
browser feature must be enabled

But, the nice part is: for most setups, both are already on by default.

So if you just want the minimal config, this is enough:

#

  browser: {
    executablePath: "/usr/bin/brave-browser",
  },
}

You only need extra browser-related config if:

you explicitly disabled the plugin before
you use plugins.allow and forgot to include "browser"
you want to force browser.enabled: true/false explicitly

So yes, if nothing else is disabling it, adding both is a bit redundant.

Docs:

https://docs.openclaw.ai/tools/browser

#

If this fixed it, please right-click the solution message → Apps → ✅ Mark Solution so it gets indexed on Answer Overflow. 🎉

kindred sun Apr 18, 2026, 10:53 PM

#

'what you mean by 'your current config is for the isolated OpenClaw-managed browser, not your already-open personal Brave session'

how can i test a grocery ordering on walmart.ca workflow executed end to end.. i seriousl y want to this recurring workflow handled by openclaw going forward... you may even want to help me a dedicated agent for this task if you think it is worth it

cedar wharfBOT Apr 18, 2026, 10:55 PM

#

What I meant is:

your current config tells OpenClaw which Brave binary to launch
it launches its own browser profile (openclaw)
that profile has its own tabs, cookies, logins, cart, etc.
so it does not automatically see your already-open personal Brave window/session

So:

managed openclaw browser = separate, safer, isolated
profile="user" / existing-session = attaches to your real logged-in browser, more powerful but higher-risk

For Walmart grocery ordering, I do think a dedicated agent is worth it, but I’d roll it out in 3 steps:

1. supervised test first

Start with the isolated browser.

#

pnpm openclaw browser start
pnpm openclaw browser open https://www.walmart.ca/

Then log in manually and tell the agent:

Use the browser to build my Walmart grocery order from my usual items. Stop before final checkout submission.

That proves the flow works without risking an accidental order.

2. create a dedicated agent

Yes, I’d make a groceries agent.

Why:

separate memory/workspace
separate standing rules

#

less chance of random chat context messing with the workflow

pnpm openclaw agents add groceries

3. add standing orders + cron

Once the supervised run works, give that agent rules like:

only use walmart.ca
build cart from saved list / prior order
flag out-of-stock items
never place order if total > your limit
always stop for approval before final submit, at least at first

Then schedule it with cron.

#

This is the right long-term shape:

browser for the site interaction
dedicated agent for isolation
standing orders for recurring rules
cron for weekly runs

Docs:

If you want, next reply I can sketch the exact groceries agent standing orders for Walmart.ca so you can paste them in directly.

kindred sun Apr 18, 2026, 10:58 PM

#

since walmart.ca would need my credential, should the openclaw use 'openclaw' profile or something else.. give me all the commands at once

cedar wharfBOT Apr 18, 2026, 10:59 PM

#

Use the openclaw profile first.

That’s the safest setup for Walmart.ca because you should log in manually in that browser and never paste your credentials into chat. Once logged in, that profile keeps its own cookies/session for future runs.

Pasteable command set

pnpm openclaw --version
pnpm openclaw gateway restart
pnpm openclaw browser --browser-profile openclaw status
pnpm openclaw browser --browser-profile openclaw start
pnpm openclaw browser --browser-profile openclaw open https://www.walmart.ca/

Then do this manually

In the OpenClaw browser window:

#

log into Walmart.ca yourself
set your store / delivery address if needed
leave that browser/profile alone

Then send your agent this prompt

Use the browser profile openclaw. Go to walmart.ca, build my grocery cart from my usual items, choose the best matching delivery or pickup slot, and stop before final order submission. Ask me before placing the order.

Optional stop command later

pnpm openclaw browser --browser-profile openclaw stop

Recommendation

Start with:

openclaw profile for testing

#how to get my agent fill form online and book tickets for me?

1. supervised test first

2. create a dedicated agent

3. add standing orders + cron

Pasteable command set

Then do this manually

Then send your agent this prompt

Optional stop command later

Recommendation