Twigg

Twigg is a hosted service that holds the context of conversations for applications using large language models. A developer creates a chat once and then sends only the next event, while Twigg keeps the history, fits it to the model in use, shortens it when it grows long, and then sends the request on. It is aimed at developers building anything from a personal agent to an enterprise application.

What Twigg offers

• Stored conversations: chat history is held outside the model providers, so only the next message is sent.
• Model routing: calls can go to Anthropic, OpenAI, Google, xAI, Fireworks or OpenRouter, and a chat can switch model part-way through.
• Context management: conversations are fitted to each model's context window, compacted or truncated when long.
• Dashboard: manages tool schemas, system prompts and context windows, and reports what each request cost.

Key facts

• Interface: REST API with bearer-key authentication and an OpenAPI specification
• Documentation: published as plain markdown for coding agents to read
• Model catalogue: live, listing the models and their prices