My AI agent is using all my credits and not completing its task

[elfin reef]

hey, I'm a new OpenClaw user and I'd like to know if it's possible to scrape Instagram profiles. I've entered some criteria, but it's getting stuck. Does anyone have a solution?

[haughty lintel]

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

[novel stump]

Yes, but Instagram is a rough target for generic agent scraping.

Most common gotchas:

• web_fetch will usually fail there, it does plain HTTP and no JS: https://docs.openclaw.ai/tools/web-fetch
• Instagram login/anti-bot often blocks sandboxed or headless automation
• automated login is a bad idea, log in manually in the OpenClaw browser instead: https://docs.openclaw.ai/tools/browser-login

Best path in OpenClaw:

1. use the browser tool, not web_fetch: https://docs.openclaw.ai/tools/browser
2. run headful, preferably host browser control
3. log in manually first
4. keep the task narrow, like one profile / one page / explicit stop condition, otherwise agents can loop and burn credits
5. if anti-bot is still the blocker, Browserbase or Firecrawl are the more realistic options: https://docs.openclaw.ai/tools/firecrawl

For your specific case, can you paste:

• openclaw --version
• openclaw status --all

• whether you're using browser, web_fetch, or Firecrawl
• the exact task/prompt you gave it

FAQ is here too: https://docs.openclaw.ai/help/faq

If you want real-world user experience in addition to troubleshooting, also ask in #users-helping-users.

[elfin reef]

So what do I need to do, how do I set it up?

[novel stump]

I’d do it in 2 stages.

Option A, recommended for Instagram
Use the browser tool, log in manually, then keep the task very narrow.

1. Make sure browser control is enabled in ~/.openclaw/openclaw.json:

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

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

2. Start the OpenClaw browser and open Instagram:

openclaw browser start
openclaw browser open https://instagram.com

3. Log in yourself in that browser.
   Do not give the model your IG credentials.
   Docs: https://docs.openclaw.ai/tools/browser-login

4. If your agent is sandboxed and needs the host browser, allow that:

{
  agents: {
    defaults: {
      sandbox: {
        mode: "non-main",
        browser: {
          allowHostControl: true
        }
      }
    }
  }
}

5. Then give it a tight prompt, for example:

Open only these 5 profile URLs, extract bio, follower count, and latest 3 post captions, then stop and return JSON.