After months of building in the open, Phase 1 of ToRun is live.

This release covers the core AI infrastructure and the day-to-day workflows most teams care about:

• Multi-provider chat — route requests across OpenAI, Anthropic/Claude, Google/Gemini, DeepSeek, xAI/Grok, Mistral, and more from a single interface
• Capability-first routing — every request flows through a pipeline that matches the mode you need to the models that actually support the required capabilities, then ranks them by price and quality; no vendor lock-in
• BYOK (bring your own key) — paste your own provider API keys into an encrypted vault; ToRun resolves them per call, scoped to whatever provider, model, or workflow you choose, with a full audit trail
• Transparent billing — every AI call writes exactly one billing record with a frozen price snapshot; invoices stay verifiable even after providers change their rates; data is exportable so you can recompute every cent offline
• Memory — long-term memory that retains durable facts across chats; user-owned, user-controllable, and language-agnostic across all 29 supported locales
• Persona foundations — AI synthetic creators with mandatory disclosure baked in before anything can post or be listed
• Generative media — image, video, and audio/music generation as first-class modes, billed per image or per second through the same routing and billing system

What is coming next

Phase 2 is in active development. No dates promised, but here is what is on deck: Council (multi-model deliberation with an adjudicator), DeepResearch with per-claim source citations, the public workflow marketplace listing flow, and the bouncyad.com ad exchange.

We will post here as each piece ships. Questions and feedback welcome in the threads below.