Context Window Calculator
Estimate how much usable context remains after system prompts, tool schemas, memory, retrieved chunks, and output reserve — before you build your RAG, MCP, or agent system.
Inputs
Total token limit for the selected model
Tokens consumed by your system/instructions prompt
Tokens used by tool/function definitions sent to the model
Tokens from prior turns kept in memory
Number of RAG/MCP chunks injected into context
Average size of each retrieved chunk in tokens
Tokens reserved for the model's response
Fill in your inputs and click Calculate to see how your context window is allocated.
Architecture Tips
- • Keep tool schemas compact — verbose schemas silently consume thousands of tokens.
- • Use sliding window or summarized memory for long conversations instead of full history.
- • Target ≤60% context utilization to leave room for unexpected response length.
- • For RAG systems, prioritize fewer high-quality chunks over many low-quality ones.
- • With MCP, each tool definition adds to your tool schema token count.