Context Mode MCP Server Cuts Claude Code Context Usage by 98%
Context Mode is an MCP server that sits between Claude Code and tool outputs, reducing context window consumption by 98%. Instead of dumping raw data into the 200K context window, it processes outputs in isolated sandboxes.
How It Works
The sandbox system spawns isolated subprocesses for each execute call. Scripts run in these subprocesses with their own process boundaries, and only stdout enters the conversation context. Raw data like log files, API responses, and snapshots never leaves the sandbox.
Ten language runtimes are available: JavaScript, TypeScript, Python, Shell, Ruby, Go, Rust, PHP, Perl, and R. Bun is auto-detected for 3-5x faster JS/TS execution. Authenticated CLIs (gh, aws, gcloud, kubectl, docker) work through credential passthrough where subprocesses inherit environment variables and config paths without exposing them to the conversation.
Knowledge Base Features
The index tool chunks markdown content by headings while keeping code blocks intact, then stores them in a SQLite FTS5 virtual table. Search uses BM25 ranking with Porter stemming applied at index time. When you call search, it returns exact code blocks with their heading hierarchy.
The fetch_and_index tool extends this to URLs: fetch, convert HTML to markdown, chunk, and index. The raw page never enters context.
Performance Benchmarks
- Playwright snapshot: 56 KB → 299 B
- GitHub issues (20): 59 KB → 1.1 KB
- Access log (500 requests): 45 KB → 155 B
- Analytics CSV (500 rows): 85 KB → 222 B
- Git log (153 commits): 11.6 KB → 107 B
- Repo research (subagent): 986 KB → 62 KB (5 calls vs 37)
Over a full session: 315 KB of raw output becomes 5.4 KB. Session time before slowdown goes from ~30 minutes to ~3 hours. Context remaining after 45 minutes: 99% instead of 60%.
Installation
Two installation methods:
- Plugin Marketplace:
/plugin marketplace add mksglu/claude-context-modethen/plugin install context-mode@claude-context-mode - MCP-only:
claude mcp add context-mode -- npx -y context-mode
After installation, restart Claude Code. Context Mode includes a PreToolUse hook that automatically routes tool outputs through the sandbox. Subagents learn to use batch_execute as their primary tool, and bash subagents get upgraded to general-purpose so they can access MCP tools.
The tool is open source under MIT license at github.com/mksglu/claude-context-mode.
📖 Read the full source: HN LLM Tools
👀 See Also
AbsolutelySkilled Registry Adds 156 Production-Ready Skills for Claude Code
A developer has created AbsolutelySkilled, a registry of 156 structured skill modules for Claude Code that persist across sessions. Each skill includes trigger conditions, reference files, test cases, and anti-patterns in SKILL.md files.
vllm-mlx fork adds tool calling and prompt cache for local AI coding agents
A developer has modified vllm-mlx to fix tool calling issues and add prompt caching, reducing TTFT from 28s to 0.3s for OpenClaw on Apple Silicon. The fork supports Qwen3-Coder-Next at 65 tok/s on M3 Ultra with working function calling.
Mozilla Thunderbolt: Open-Source Enterprise AI Client for Self-Hosted Infrastructure
Mozilla announced Thunderbolt, an open-source AI client under MPL 2.0 license designed for organizations to deploy self-hosted AI infrastructure with model choice, enterprise data integration, and cross-platform native applications.
Manual-Driven Development: A Method to Prevent Claude Code's Confident Divergence
Manual-Driven Development (MDD) is a method that addresses confident divergence in Claude Code, where the AI produces wrong code that passes its own tests. In a production audit, MDD found 190 issues, wrote 876 new tests in under 8 hours, and eliminated rule violations.