A Claude Code plugin that checks every file and shell call an AI agent makes, fixes what it safely can, and tells the agent how to fix the rest. Windows and macOS.
Your AI agent sends a bad command. io-guard fixes it before it runs, and the agent gets the result it wanted.
The agent keeps working on your code. It doesn't lose time, and it doesn't spend tokens trying again and again to fix errors that come from the shell, not from your code.
io-guard is a plugin for Claude Code. It works on Windows and macOS. It looks at every command and every file change the agent makes:
- Before it runs: when io-guard can fix it, it fixes it, and tells the agent what it changed. - When it can't be fixed: io-guard stops it, and tells the agent what to send instead. - After it runs: io-guard tells the agent what really happened, in clear words, not just "exit code 1".
You can change every rule and every limit, for all your projects or for one.
Agents make the same shell and file mistakes every developer knows. io-guard fixes them before they run:
The agent sends Without io-guard What runs instead --------- A long Python script in a heredoc, on Windows unexpected EOF while looking for matching ''. The Bash tool cuts any command near 7.8 KB The script, read from a file, exactly as written print("C:\\work\\app") in a heredoc, on Windows The Bash tool halves the backslashes. Python reads \a as a bell character and prints C:\work and junk The script, read from a file, every backslash kept cd "C:\work\app\" unexpected EOF. In bash, \" is a quote inside the string, not its end cd "C:/work/app/" taskkill /F /IM app.exe in Git Bash Git Bash turns /F into the path F:/, and taskkill fails The same command, with /F and /IM left alone dir 2nul in Git Bash A
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/Riztazz--claude-io-guard.html)
Find which APIs of your pinned Python dependencies changed after your coding model's training cutoff, and give the agent short AGENTS.md / CLAUDE.md notes from a static API diff. No model calls, no AP
00200200/usagetrimKeep the signal. Cut the noise. Local CLI + MCP — fold verbose tool output for Claude Code, Codex, Cursor & Desktop.
dontpayfull/AInotateAnnotated screenshots for AI agents: Skitch-style arrows, numbered steps, boxes, labels, solid redaction
dongnguyenvie/bashcut-agent-kitBashCut's agent kit: editing skills (footage survey, beat cuts, audio mix, captions, colour, voiceover…) and the BashCut MCP server for Claude Code and Codex.
Morningstar202604/AgentSeedAnti-hallucination guardrails for AI coding agents - hybrid Skill + MCP server (6 tools) that verifies code before it is marked done. Works with Claude Code, Cursor, VS Code, Copilot.
luckeyfaraday/frontier-orchestratorMulti-agent orchestration for Claude Code: MCP server + skill that keeps Claude as lead engineer while delegating backend work to OpenAI Codex and frontend/design work to Kimi
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.