Context Mode

Active

GitHub TypeScript NOASSERTION

Description

Context Mode is a context window optimization tool for AI coding agents that sandboxes tool output for 98% context reduction across 12 major platforms.

Key Features

Sandboxes tool output, reducing 315KB raw data to 5.4KB — a 98% context reduction
SQLite + FTS5 session continuity that recovers key context via BM25 search after conversation compaction
Think in Code paradigm: LLM writes analysis scripts instead of processing raw data, saving 100x context
Supports 16 major AI coding platforms including Claude Code, Cursor, Codex CLI, and more
Provides 11 MCP tools including ctx_execute, ctx_search, ctx_doctor for indexing, search, and diagnostics

Use Cases

💡 Preventing context window exhaustion during long AI coding agent sessions

💡 Achieving cross-session task continuity in tools like Claude Code

💡 Optimizing context consumption from raw data generated by MCP tool calls

💡 Reducing token consumption costs for AI coding assistants in enterprise development environments

Strengths & Limitations

✅ Strengths

• Actively maintained, recent updates
• High community interest (20.3k stars)

⚠️ Limitations

• No clear open-source license

Quick Start

Claude Code install: run /plugin marketplace add mksglu/context-mode, then /plugin install context-mode@context-mode, restart and verify with /context-mode:ctx-doctor; also available via npm and VS Code Marketplace. Use /context-mode:ctx-execute for sandboxed code execution, /context-mode:ctx-stats to view context savings.

Visit GitHub Visit Website View Docs

Related Projects

Claude Context

12.5k · TypeScript

Normal A

Code search MCP for Claude Code and coding agents. Makes entire codebases available as context for AI coding assistants using vector-based semantic code search for precise understanding of large projects.

mcptypescriptcoding +3

· MCP plugin providing semantic code search across entire codebases for Claude Code and AI coding agents
· Vector database-backed retrieval using Milvus/Zilliz Cloud for efficient large-codebase indexing
· Multi-client support: Claude Code, Cursor, Windsurf, Gemini CLI, Codex CLI, Qwen Code, Void, and Claude Desktop

View full details → View on GitHub

HumanLayer

11.4k · TypeScript

Normal B

HumanLayer provides a human-in-the-loop layer for AI coding agents, enabling them to seek human approval and guidance when solving hard problems in complex codebases.

human-in-the-loopcodingclaude-code +2

· Open-source IDE (CodeLayer) for orchestrating AI coding agents in complex codebases
· Keyboard-first workflows with battle-tested patterns for builders who value speed
· Advanced context engineering — scale AI-first development to entire teams

View full details → View on GitHub

CodeGraphContext

4.2k · Python

Active A

CodeGraphContext is an MCP server that indexes local codebases into a graph database, providing structural context to AI coding assistants for precise code understanding and navigation.

mcpcode-graphcode-analysis +2

· MCP server and CLI that indexes code repos into a queryable graph database
· Supports parsing and analysis for 22 programming languages
· Query callers, callees, class hierarchies, and call chains

View on GitHub

skybridge

2.0k · TypeScript

Active A

Skybridge is a full-stack TypeScript framework for MCP Apps and ChatGPT Apps. Type-safe. React-powered. Platform-agnostic.

agentaiapps-sdk +5

· Full-stack React framework for building type-safe MCP Apps and ChatGPT Apps
· Write-once-run-everywhere — abstracts differences between Claude, ChatGPT, and VSCode
· Dev server with local emulator, HMR, and permanent tunnel for live testing

View on GitHub

上下文工程长上下文RAG

Context Engineering: Context Decay and Recovery in Long-Conversation Agents

Long-conversation agents fail at context management, not model capability. A systematic comparison of sliding window, retrieval injection, and layered compression strategies with practical decay diagnosis and recovery patterns.