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
ARP: Stateless WebSocket Relay for Autonomous Agent Communication
ARP (Agent Relay Protocol) is a stateless WebSocket relay for autonomous agent communication featuring Ed25519 identity, HPKE encryption per RFC 9180, binary TLV framing, and 33 bytes overhead per message. No accounts or registration required—just generate a keypair and connect.
Nakkas MCP Server Generates Animated SVGs from AI Descriptions
Nakkas is an MCP server where AI constructs complete animated SVG configurations from descriptions, rendering clean animated SVGs with shapes, gradients, animations, and filters. It supports parametric curves, 15 filter presets, CSS @keyframes and SMIL animations, and works anywhere SVG renders.
Open Source Curated Collection of OpenClaw Resources Unveiled
Discover a new open-source collection of OpenClaw resources, curated by the community to enhance AI development and collaboration.
Claude-rank: Claude Code Plugin for AI Search Visibility Audits
Claude-rank is a free Claude Code plugin and CLI that audits technical foundations for AI search visibility, handling technical SEO, AI citability scoring, crawlability checks for AI bots, and automated fixes for discoverability issues.