Breaking

Open-Source Plugin Slashes AI Coding Token Usage by 98%, Extending Sessions From 30 Minutes to 3 Hours

October 10, 2026 at 10:15 EDT

Software Dev & Coding
Open Source

A new MCP server called Context Mode dramatically reduces token consumption in AI coding workflows by intercepting tool outputs before they flood the context window, compressing 315 kilobytes of raw data down to 5.4 kilobytes while preserving searchable session memory.

AI Developer Tools

Context Mode Eliminates 98% of Token Waste

Open-source plugin extends AI coding sessions from 30 minutes to 3 hours by sandboxing large tool outputs before they consume the context window.

98%

Token compression in raw outputs

6×

Session duration extension (30 min → 3 hr)

26.1k

GitHub stars

17+

Supported platforms

Token Compression Examples

56KB

Playwright
snapshot

299B

Compressed

How it works

Instead of dumping large tool outputs directly into context, Context Mode routes them through isolated subprocess sandboxes, maintaining a searchable SQLite database while returning only compressed summaries to the conversation window.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.