Air
Shrink noisy tool output from read, grep, diff, bash, tests, web, and search before it fills the agent context window.
Overview
Air is a MCP server for the Build phase that compresses read, grep, diff, bash, test, web, and search tool output for leaner agent context.
What is this MCP server?
- AI-oriented compression across read, grep, diff, bash, test, web, and search outputs
- npm package @10iii/air-mcp-server v0.2.8 with npx runtime hint
- stdio MCP transport for Claude Code, Cursor, and similar clients
- Reduces token burn on large logs, diffs, and command stdout
- GitHub repo 10iii/air, subfolder packages/mcp-server
- Version 0.2.8
- Covers read, grep, diff, bash, test, web, search categories per description
- stdio transport via npm @10iii/air-mcp-server
What problem does it solve?
Raw terminal and repo tool output routinely exceeds what an LLM can use, wasting tokens and obscuring the one line that matters.
Who is it for?
Solo devs on Claude Code or Cursor running repetitive grep, test, and diff loops on medium or large repos.
Skip if: Workflows that only need a single small file read with no log or search volume.
What do I get? / Deliverables
Agents receive condensed, task-relevant views of logs and files so more steps fit in one session.
- Registered Air MCP server v0.2.8
- Compressed views of multi-tool dev output
- Lower average tokens per read/grep/bash/test cycle
Recommended MCP Servers
Journey fit
How it compares
Output compression MCP for dev tools, not a browser automation or deployment server.
Common Questions / FAQ
Who is Air for?
Builders using MCP-enabled coding agents who routinely process large diffs, test logs, or search results.
When should I use Air?
When context fills up from bash, test, grep, or web tool calls and you want structured truncation instead of manual copy-paste.
How do I add Air to my agent?
Register stdio MCP with npx @10iii/air-mcp-server per the server schema in the 10iii/air repository.