Claude Tool Output Compressor

Claude Tool Output Compressor MCP Connector for Claude

A+

Reduces context window exhaustion by applying deterministic compression rules to large tool outputs.

3 tools Official Updated Oct 1, 2026 Official Vinkius Partner

This MCP server prevents context window exhaustion by applying deterministic, tool-specific compression rules to large outputs. Instead of simple truncation, it uses 'Smart Reduction' to preserve critical information. For example, it uses compress_output to keep error lines in Bash outputs, filter logs for warnings, and prune empty fields in API responses. It is designed to maintain the signal-to-noise ratio for AI agents using Cursor, VS Code, Claude Desktop, and Windsurf.

compressioncontext-windowtoken-optimizationllm-efficiencydeterministic

3 tools expose this connector's capabilities to your AI agent.

compress_output

estimate_compression_gain

get_compression_statistics

See how to talk to your AI agent using Claude Tool Output Compressor.

Compress this large bash output for me.

The output has been compressed. I have kept the first 20 lines, the last 20 lines, and all lines containing ERROR to ensure you have the necessary context.

How much space can I save by compressing this API response?

The estimated reduction is 45%, which will bring the token count down to approximately 150 tokens.

Show me the current compression efficiency stats.

Total tokens saved: 12,450. Average compression ratio: 0.35. Total operations performed: 12.

The server applies deterministic rules based on the source tool. For instance, Bash outputs retain the start, end, and error lines, while API responses are pruned of null values and large arrays.

Related Connectors