FastRecall is a provider-independent context API for applications that move conversations across AI models. It stores ordered history once and returns model-ready context without proxying model calls, helping routers and multi-agent systems keep continuity while retaining control of provider keys.

What's included in this release:

• Applications can store conversation history once and recall it for direct use with OpenAI, Anthropic, Gemini, and other providers.
• Free, low-latency recalls separate retrieval costs from the metered storage used by each context.
• Replay mode returns an assembled prompt, while native continuation preserves supported provider caching behavior.
• Branches let developers fork a context from a chosen point without duplicating or rewriting the original history.
• FlashCompact organizes contexts beyond one million tokens with model-free compaction when full-fidelity replay is unnecessary.

Together, portable storage, two recall modes, branching, and compaction give developers a practical context layer that can follow an application across models.