Constraints-first
Define success criteria and guardrails before running tools.
Parley Labs is building a product-grade agent loop: plan → search → execute → verify — with artifacts, diffs, and policy gates.
Minimal surface area. Maximum clarity. Every action is inspectable.
Define success criteria and guardrails before running tools.
Every run yields diffs, logs, images, and outputs—replayable.
Permissioned tools with audit logs and explicit user approvals.
Type anything you want — we only store name + email on the backend.
Invite-only while we harden runtimes, security, and tooling.
No. After initial setup and model downloads, ParleyAI works completely offline.
Yes. We designed ParleyAI with zero network calls during normal operation. You can verify this yourself — the app works with all internet blocked.
Most alternatives rely on cloud connections and monthly subscriptions. ParleyAI runs 100% locally with a one-time purchase (plus optional add-ons later).
Only name and email. The chat UI can accept any text, but the backend ignores everything except those two fields.
Server-side validation + rate limiting. Requests must be valid email format, limited by IP, and we only accept the two whitelisted fields.
To your backend endpoint (recommended). This package includes ready-to-deploy backend examples that deliver leads to noreply@parley.live via authenticated SMTP.