MCP Server
aict ships a standalone MCP (Model Context Protocol) server binary so AI assistants can call every tool directly — no shell spawning, no output parsing.
Build
go build -o aict-mcp ./cmd/mcp
Configure Claude Desktop
Add to ~/.config/claude/claude_desktop_config.json:
{
"mcpServers": {
"aict": {
"command": "aict-mcp",
"args": []
}
}
}
Configure Cursor
Add to .cursor/mcp.json in your project root:
{
"mcpServers": {
"aict": {
"command": "aict-mcp",
"args": []
}
}
}
Available Tools
Once connected, every aict tool becomes a typed, callable function:
| Tool | Description |
|---|---|
ls | Directory listing |
grep | Pattern search |
cat | File read |
find | Filesystem search |
stat | File metadata |
wc | Line/word/byte count |
diff | File comparison |
git status | Git status |
git diff | Git diff |
git log | Git log |
git ls-files | Git tracked files |
git blame | Git blame |
checksums | Hash computation |
ps | Process list |
env | Environment variables |
system | System info |
doctor | Self-diagnostic |
How It Works
- The MCP server starts and registers all tool specs
- The AI client (Claude, Cursor) discovers available tools
- When the model decides to call a tool, it sends a structured request
- The server runs the tool, captures structured JSON output, and returns it
- The model receives typed data — not raw terminal output
Troubleshooting
| Issue | Fix |
|---|---|
| Server won’t start | Ensure aict-mcp is in your PATH |
| Tools not showing | Restart your AI client after config change |
| Permission errors | Check file permissions on target directories |