Uni-CLI — Governed Software Commands + MCP for Agents

Main

• Treat it as your agent substrate: search → execute → inspect evidence. Keep the outputs machine-readable (-f json) for downstream agents.

• Start with low-risk sites/commands and a restrictive permission profile; then expand coverage once you have auditing in place.

• Use MCP serve mode when you want to expose the catalog as tools to your agent runtime instead of shelling out directly.

• Lean on “repair” workflows: README highlights adapter repair paths so broken integrations can be fixed rather than silently failing.

Source-backed notes

• README Start In 30 Seconds shows npm install -g @zenalexa/unicli, unicli search ..., and npx @zenalexa/unicli mcp serve.
• README badge line lists quantitative counts (sites/commands/pipeline steps/tests) embedded in the doc.
• README includes protocol entry points and examples for MCP serve and other modes.

FAQ

• Do I have to run it via MCP?: No — README shows you can run commands directly with unicli ..., then add MCP serve when needed.
• Why include counts like sites/commands?: They help you gauge coverage; README embeds site and command counts in the header.
• How do I keep executions safe?: Start with restrictive permission profiles and prefer evidence-rich outputs for auditing.