AI Pipeline Tool

Context Manager

Manage your AI context window efficiently. Estimate tokens, plan budgets, and optimize what stays in context for peak LLM performance.

Install

Get the CLI Tool

Run the context manager locally as an MCP server, or try the interactive demo below.

npx @clinetools/context-mgr
Requires Node.js 18+
  • Token estimation — approximate counts for code and prose
  • Priority-based context planning with 3 strategies
  • Smart trimming — truncate, summarize, or drop blocks
  • Actionable recommendations for context optimization
  • Zero config — just run with npx
Tutorial

How to Use It

Three ways to manage your AI context window — pick the one that fits your workflow.

1

Try Online

Use the interactive demo below to add content blocks and plan your context budget — no install needed.

↓ Jump to interactive demo
2

Use via CLI

Run as a local MCP server and connect any MCP-compatible client.

npx @clinetools/context-mgr
3

Add to Cline / Claude Code

Add the tool to your MCP settings for context management from your AI assistant.

"context-mgr": { "command": "npx", "args": ["@clinetools/context-mgr"] }

MCP Client Configuration (Cline) 

{
  "mcpServers": {
    "context-mgr": {
      "command": "npx",
      "args": ["@clinetools/context-mgr"]
    }
  }
}

Claude Code Configuration 

# In your project's .mcp.json:
{
  "mcpServers": {
    "context-mgr": {
      "command": "npx",
      "args": ["@clinetools/context-mgr"]
    }
  }
}

Example Prompt: Estimate Token Usage 

// Prompt to your AI agent:
"Estimate how many tokens this file uses"

// The agent calls:
estimate_tokens({
  text: "...file contents..."
})
// Returns: { tokens: 1250, characters: 5000, words: 820 }

Example Prompt: Plan Context Window 

// Prompt to your AI agent:
"I have 8K tokens left. Plan which files and
conversation history to keep in context."

// The agent calls:
plan_context({
  blocks: '[{"id":"system","content":"...","priority":"critical","category":"system"}, ...]',
  budgetTokens: 8192,
  strategy: "balanced"
})
Live Demo

Try It Online

Add content blocks, set a token budget, and see how the context planner decides what to keep.

Context Window Planner

Add content blocks and plan your context budget

Load scenario:
8K

Add content blocks and click Plan Context to see the optimization plan.

Context Plan

Reference

Understanding Context Management

Key concepts for managing LLM context windows effectively.

Context Window Limits

Every LLM has a maximum context length — from 4K to 200K tokens. Exceeding it causes truncation or errors. Know your model's limit: GPT-4o has 128K, Claude has 200K, Gemini has 1M.

Token Estimation

Tokens are not words. Code produces more tokens per character than prose due to symbols and operators. A rough estimate: ~4 chars/token for English, ~3.5 chars/token for code.

Priority-Based Trimming

Not all context is equal. System prompts are critical, recent conversation is high-priority, and old reference material can often be dropped. Smart trimming keeps the most valuable content.

Conversation Compression

Long conversations accumulate tokens fast. Strategies: summarize older turns, drop low-value exchanges, keep only the most recent N turns, or compress reference material into bullet points.

Manage Every Context

Our Pro plan includes automated context management across all your AI pipelines.

View Plans