Weavin

Weavin's pitch is honest about its shape: it is not a platform you build on, it is a worker you hand a job to. The four-line summary on the homepage — say it once, it's done every day; it asks before spending or sending — is the whole product, and that focus is what makes it credible. Where it differs from a generic agent runner is the approval boundary. Anything that costs money or is irreversible lands in a 'Needs you' list; everything else the agent does quietly and logs. For a shop owner or a solo dev mid-launch, that inversion — you only show up at the last step — is the value, not the automation itself. The strengths: plain-language task entry (no flows, no rule syntax), bring-your-own-key for live calls on Anthropic, OpenAI, Google, or OpenRouter, per-business avatars with separate memory and boards, and a credit system that stops rather than auto-charges when you run out. The docs confirm no coding and no deployment step — you describe your business, review the profile it drafts, connect tools, and launch the same sitting. The weaknesses are real and worth stating. It's in invite-based beta, so access is gated. Scheduled runs never use your own key — they bill platform credits regardless, which is the single biggest cost gotcha. There is no multi-agent orchestration, no role-based approval hierarchy, no on-prem option, and no deep model fine-tuning. The integrations list is thin by design: it connects to what you already use, but don't expect a marketplace of hundreds of pre-built apps. And per-credit costs climb fast if your scheduled workload is heavy. Where it fits: solo operators with a repeating, judgment-light, drafting-heavy workload — customer replies, reorder sheets, DM triage, launch-post follow-up. Where it doesn't: teams that need approval chains, orgs with compliance requirements that demand self-hosting, and anyone whose workload is mostly high-volume scheduled runs where platform credits become the dominant expense.