Everything you need for trustworthy AI agency
Custom browser with OAuth3 token management, recipe execution, and evidence capture. Three surfaces: web automation, machine access, and cloud tunnel.
Terminal-first orchestration. Manage skills, run recipes, and coordinate agents from the command line. Extends Stillwater CLI with OAuth3 vault and cloud sync.
Curated collection of verified AI agent workflows. Gmail triage, Slack digest, LinkedIn outreach, and more. Each app tested to rung 65537 before shipping.
Fine-grained permissions: gmail.read.inbox, machine.execute.shell, tunnel.connect. Every scope is time-bounded, revocable, and auditable.
Action budgets (emails/day, posts/day), spend budgets (transaction caps, daily limits), and recipient allowlists. Fail-closed if any gate fails.
High-risk actions (delete, send, spend) always require human sign-off. Preview → Cooldown → Approval → Execution. No shortcuts.
Every state transition produces a SHA-256 hash linking to the previous event. Tamper-evident by construction. Append-only evidence chains.
Mode A: Screenshots + action log. Mode B: Full page archives + DOM. Mode C: Full archives + PZip compression (Part 11 ready).
Sealed approvals, edit tracking with override reasons, electronic signatures, and 90-day retention. Built for regulated industries.
First run uses LLM to discover the workflow. Subsequent runs replay the recipe at 99% lower cost. Recipes accumulate as platform capital.
Users buy outcomes, not models. L1 (fast) through L5 (adversarial). Routable across providers. Swap engines without user disruption.
Page-aware chat helper. Summarizes pages, suggests clicks, builds recipes, and tracks credits. Available on app store, browser, and documentation pages.
