billion-context is a context-compression proxy and plugin for AI coding agents that sits between the agent and its Anthropic- or OpenAI-compatible model API.
Follow billion-context to get an email when its listing changes, such as new pricing or features. One digest a day at most, and you can unsubscribe anytime.
It injects context-management tools so the model itself decides when and what to fold into layered, reversible summaries, keeping long sessions inside a small context window while preserving the prompt-cache prefix.
- 01
Model-driven incremental compression via injected compress, decompress, search_context and acp_status tools
- 02
Reversible, prefix-cache-friendly summaries targeting ~5x fewer tokens and month-long single sessions within a 100K window
- 03
Launchers and native plugins via the `bili` CLI (`bili claude`, `bili codex`, `bili opencode`, `bili pi`, `bili plugin install <host>`)
- 04
Broad host support: Claude Code, Codex, OpenCode, pi, Gemini CLI, Qwen Code, Kimi, Hermes, Aider, GitHub Copilot CLI, Amp and more
- 05
Built on the open acp-kernel compression engine with an accompanying open preprint and cache-health diagnostics (`/acp`, `/acp-cache`)
Free
Free and open source under the MIT License with an added attribution term for products that incorporate it. You keep paying your own model provider; the proxy aims to reduce token usage rather than add fees.
Best for
Developers running very long coding-agent sessions who hit context-window limits or high token bills and want compression that the model controls and can undo.
Not ideal for
Users who cannot route agent traffic through a local proxy or certificate MITM for closed-binary hosts, or teams whose products cannot meet the license's visible-attribution requirement.
An ambitious, actively maintained context-compression layer with unusually broad agent coverage and a published methodology; worth trying for marathon sessions, with the caveats of proxy setup and an attribution-bearing license.
Ships frequent releases (v0.1.197 on October 14, 2026) and trended on GitHub's weekly chart in mid-October 2026; recent work expanded native plugins for hosts such as Kimi Code, Hermes and zcode.
As of
Is billion-context free?+
Yes - billion-context is Free. Free and open source under the MIT License with an added attribution term for products that incorporate it. You keep paying your own model provider; the proxy aims to reduce token usage rather than add fees.
Is billion-context open source?+
billion-context is not listed as open source on ai.dosa.dev - it is Free.
Who is billion-context best for?+
Developers running very long coding-agent sessions who hit context-window limits or high token bills and want compression that the model controls and can undo.
Who is billion-context not ideal for?+
Users who cannot route agent traffic through a local proxy or certificate MITM for closed-binary hosts, or teams whose products cannot meet the license's visible-attribution requirement.
What are the best billion-context alternatives?+
The closest billion-context alternatives on ai.dosa.dev are Pieces for Developers, agenttrace, Vibe Kanban - all listed under Developer Productivity & Workflow.
Who makes billion-context?+
billion-context is developed by ranxianglei. It is listed in the Developer Productivity & Workflow category on ai.dosa.dev.
Favorite this tool to revisit it later, or Zap it to contribute to the public vote count.