Twigg

Twigg is a hosted context layer for applications that call large language models. Instead of rebuilding and resending an entire conversation on every request, a developer creates a chat once and sends only the next event; Twigg stores the conversation outside the model providers, fits it to whatever model is being called, compacts or truncates it when it runs long, and routes the call. Its documentation lists Anthropic, OpenAI, Google, xAI, Fireworks and OpenRouter as destinations, and because the conversation lives in Twigg a chat can switch model part-way through. Tool schemas, system prompts and context windows are managed from a dashboard, which also reports what each request cost. The API is a REST service under api.twigg.ai authenticated with a bearer key, with an OpenAPI specification and the whole documentation set published as plain markdown for coding agents to read. Model prices are published in a live catalogue that states the platform markup included in them, and the catalogue price is frozen when a run opens so a later change is never applied retroactively; the company publishes no subscription price. Twigg positions itself for anything from a personal agent to an enterprise application, and this was its second Product Hunt launch.