OpenClaw burned tokens and failed a basic email login
Richard tested non-vendor-lock-in alternatives to Grok Bot and was blunt about OpenClaw. He calls the software poorly designed, says it will burn all your tokens unless you hire another AI to keep reconfiguring it, and says it failed basic tasks like logging into email and sending one message. That matches a pattern we already see in OpenClaw posts: powerful when tuned, brittle when left alone. If you try it, start with one narrow job, set spending caps on the model keys, and do not hand it inbox login until a dry run works.
The lesson: OpenClaw can burn model tokens and still fail a basic email login unless you keep reconfiguring it.
Copy this prompt
Before connecting OpenClaw to my email, run a dry test: read one sample message I paste, draft a one-line reply, and stop. Do not store my password. Set a hard token or dollar cap of [LIMIT] per day. Only after I approve the draft may we try a real send on a throwaway account.
Our own prompt, written from what the source describes. We haven't run it ourselves. Text in [BRACKETS] is what you replace.
Watch out for
- Set API spending caps before the first long run. Token burn is a common complaint.
- Do not start with production email. Prove a draft-only flow first.
- Plan time to maintain it. Several users say it needs ongoing reconfiguration.
Where this comes from
Original post by @odwyer_richard on X, Nov 5, 2026. We rewrote it in our own words. Check the original for the full context.
Read the originalWant to do this yourself?
Our guide compares which agents can do this job, with steps and a prompt.
Turn a messy inbox into a 5-minute daily summary
Paste in today's emails and get back what needs a reply today, what can wait, what to ignore, and a one-line reply for each. It also flags emails that look like scams.