# Twigg

> Twigg is a context store, assembler and model router in one: a single API for LLM use. Create a chat, then send only what just happened: a prompt or a tool result, plus the model you want. Twigg keeps the context, fits it to that model's window with per-namespace budgets and retention, translates it for Anthropic, OpenAI, Google, xAI, Fireworks or OpenRouter, streams the answer back as server-sent events, and records what it sent and what it cost. A chat started on one model can be continued on another by changing one field. Your application keeps the agent loop and executes the tools.

Base URL `https://api.twigg.ai`, one API key as a bearer token, nine routes under `/api/v1`. No SDK.

## Docs

- [Quickstart](https://twigg.ai/docs/quickstart.md): create a chat, send a turn, read the stream, answer a tool call, error codes.
- [Overview](https://twigg.ai/docs/overview.md): concepts, namespaces, authentication, the loop, the endpoint table.
- [Context control](https://twigg.ai/docs/context.md): context budgets, retention policies, compaction, where a policy lives.
- [API reference](https://twigg.ai/docs/api.md): every endpoint, request and response shape, the stream event vocabulary.

## Models

- [Model catalogue](https://twigg.ai/v1/catalogue/models.md): every model you can name on a run, with context window, capabilities and prices per million tokens including the platform margin. Generated live from the catalogue.

## Optional

- [Everything in one file](https://twigg.ai/llms-full.txt)
- [Console](https://twigg.ai/dashboard): API keys, usage, billing and configuration for a signed-in organisation.