Refund RaiderConsumer action, with teethTalk to Refund Raider

Voice-first consumer action agent

Refund Raider

Talk to an agent that can fix refunds, cancellations, replacements, and billing fights.

Start with voice or chat. Refund Raider turns messy merchant problems into a live action case with evidence, approvals, execution, and follow-up instead of another support script the user has to finish alone.

Input

Live voice/chat, Gmail, screenshots, merchant URLs, uploaded proof, and a companion browser runtime for logged-in tasks.

Output

Policy evidence, action plan, approvals, execution, and follow-up timeline.

Stack

Firecrawl Search, ElevenLabs, Supabase, Resend.

Live agent flowTalk → plan → pair → execute

Action ready

“I canceled the trial, still got charged, and want the fastest path to fix it.”

Evidence

Billing path and cancellation evidence line up.

Policy, account clues, and prior messages are enough to choose the first action path and line up the follow-up fallback.

Plan

Primary channel, consent gate, and follow-up queue

Refund Raider stops at the exact user-approval edge, then runs the action and keeps the case alive after the first send or paired browser step.

Built to grow from a sharp demo into a real startup product.

Open the live intake →

Why voice works

The story lands fast because the agent is the shell, not a decorative side panel.

No dashboard mosaic. No noisy SaaS chrome. Just conversation, evidence, plan, approvals, execution, and a clean follow-up state.

Talk first

Start with the live agent, not a form. Describe the problem in plain English and let the case take shape from there.

Prove the case

Every recommendation is backed by visible policy evidence so the user sees why the path should work.

Run the plan

Refund Raider drafts, queues approvals, runs the strongest action path, and tracks the follow-up state.

Startup flow

Show the agent first, then prove it with one real action case.

  1. Step 1

    Start talking: tell Refund Raider what you want fixed, canceled, replaced, or refunded.

  2. Step 2

    The agent extracts the facts, finds policy evidence, and builds an executable action plan.

  3. Step 3

    For account-bound flows, pair the companion browser runtime and let the agent drive the real support session.