# Context Controller

> Context window management for AI agents: smart compaction, priority-based
> eviction, and semantic recall. Run longer agent sessions without hitting
> context limits.

## Pages

- [Home](https://contextcontroller.com/): product positioning, features overview, how it works, testimonials.
- [Live demo](https://contextcontroller.com/demo.html): interactive client-side context-window visualizer — pick a model size (8K/32K/128K/1M), add messages, watch usage, trigger compaction.
- [Features](https://contextcontroller.com/features.html): compaction policies, priority tiers, semantic recall, multi-agent shared context, observability.
- [Pricing](https://contextcontroller.com/pricing.html): Starter (free), Pro ($29/seat/mo), Team ($99/seat/mo); monthly/annual billing.
- [Docs](https://contextcontroller.com/docs.html): quickstart, Python + JavaScript SDK examples, REST API reference, glossary.
- [About](https://contextcontroller.com/about.html): mission, beliefs, team.
- [Contact](https://contextcontroller.com/contact.html): contact form (front-end only).

## Key concepts

- **Compaction**: summarizing verbose content in place to reclaim tokens.
- **Priority tiers**: pinned / high / normal / low — eviction order under memory pressure.
- **Semantic recall**: retrieving evicted content on demand via embeddings.
- **Managed token**: any token processed by the compaction engine; the billing unit.

## Integration

Drop-in proxy for any OpenAI-compatible API, or wrap clients with the Python /
JavaScript SDKs. See the docs quickstart for code examples.