I will build an autonomous ai agent loop that runs your task daily on claude


About this gig
Your AI doesn't need better prompts. It needs a finish line.
Most AI automations fail the same way: nobody defined what DONE looks like, so the agent guesses, stops too early, or runs forever. I fix that. I'm a systems developer with 50+ years of software experience, and I build loops, not prompts - using the loop-engineering method now standard in Claude Code.
Every loop I deliver has:
- A validated goal: a clear end state your AI can PROVE it reached (a count, a check, a test - not "looks good")
- - Guardrails: plain-word limits ("never delete", "never reply") plus a hard turn cap so it can't run away
- - A timer: a cloud routine that runs it daily or weekly with your laptop closed
- - An independent checker: a second agent reviews the work; no AI grading its own homework
- - Read-only first: it summarizes for days before it's allowed to touch anything
Real examples: inbox triage that labels what matters every morning. Monthly report first-drafts. Support tickets closed - and re-opened by the checker if closed wrongly.
You get documentation, a live demo, and revisions until your loop runs clean.
Tell me the task you repeat every week. I'll hand it a finish line.
Get to know Reginaldgpt
AI chatbots custom GPT systems that automate your business 24 7
- FromHaiti
- Member sinceJan 2023
- Avg. response time8 hours
- Last delivery4 months
Languages
English, French
FAQ
What do I need before ordering?
A paid Claude plan (routines are a paid feature) and the accounts your loop touches (Gmail, Slack, etc.). For terminal /goal loops, Claude Code v2.1.139+. I confirm everything in a pre-order checklist - if your plan can't support it, I tell you before you pay.
Is my data safe? Do you need my passwords?
No passwords. Connectors are authorized on YOUR side in Claude settings; API keys go into your routine's environment variables, never shared with me. Every loop starts read-only - it summarizes for days before it's allowed to change anything.
What if the AI makes a mistake?
That's why the checker exists. Premium loops include a second, independent agent that reviews the doer's work and can undo it - re-open a ticket, un-archive an email - and escalate to you. Every loop also has hard limits and a turn cap so it can never run away.
What tasks work best?
Anything repeated with a countable done: inbox triage, file organizing, report first-drafts, spreadsheet labeling, caption cleanup, support tickets, news monitoring. If done can't be counted or tested, I tell you upfront and propose the nearest verifiable version.
Do you offer ongoing support?
Premium includes 30 days. After that, the monthly maintenance extra covers tuning, new guardrails, connector changes, and one new mini-loop per month. Loops are living systems - the retainer keeps them sharp as your workflow changes.
Can you work in French or Haitian Creole?
Oui / Wi. I'm fully bilingual and deliver documentation in English, French, or Haitian Creole. I specialize in serving Haitian organizations and francophone businesses that other AI freelancers can't.

