A sales agent can draft emails quickly; the harder problem is deciding who to contact, checking that an address is deliverable, and preventing a bad message from going out unattended. This project treats outreach as a guarded pipeline, with human approval as the default rather than an afterthought. Its subscription-based model access also avoids separate per-token API billing, though it still depends on eligible Claude or ChatGPT plans and a sending mailbox.
What Sets It Apart
- Prospecting is tied to evidence on each business’s own website. It looks for buying signals and named decision-makers, helping ground outreach in specific details rather than generic list-based personalization.
- Email checks distinguish certainty from guesses. Addresses are labeled as verified, risky, guess, or invalid, and only deliverable addresses are eligible to send. This makes uncertainty visible instead of treating a guessed address as a successful match.
- The workflow includes reply handling and opt-outs. It can draft responses and set aside messages that need human judgment, so the automation covers more than just sending the first email.
- Approval and sending limits are built into the workflow. Users can review messages before sending, while daily caps and recipient office hours help constrain the system’s reach.
Who It’s For
Great fit if you want to run a small, self-hosted outbound-sales operation and are comfortable managing a mailbox, prospecting rules, and deliverability obligations. Look elsewhere if you need a managed sales platform, large-scale contact data, or fully autonomous sending: it relies on subscription CLI access, a separate email account, and careful oversight of cold outreach.
