Agent Launch banner
dhruv-anand-aintech dhruv-anand-aintech

Agent Launch

Development community

Description

One normalized launcher for Codex, Claude Code, Gemini CLI, OpenCode, and Cursor Agent

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

README

agent-launch

![agent-launch — unified CLI for coding agents](docs/assets/agent-launch-hero.webp)

One command for launching the local coding-agent CLIs installed on this machine:

  • Antigravity CLI (agy)
  • Aider (aider)
  • Amp (amp)
  • Codex CLI (codex)
  • Claude Code (claude)
  • Cline (cline)
  • Factory Droid (droid)
  • Gemini CLI (gemini)
  • Grok Build (grok)
  • Kilo Code (kilo)
  • Kimi Code (kimi)
  • Kiro CLI (kiro-cli, exposed as --agent kiro)
  • Kiro Crew (kirocrew, exposed as --agent kirocrew)
  • Muse Code (muse)
  • OpenCode (opencode)
  • Pi Coding Agent (pi)
  • Pier Code (pier)
  • Qwen Code (qwen)
  • Trae Agent (trae-cli, exposed as --agent trae)
  • Cursor Agent (agent, exposed as --agent cursor)
  • Random selection across installed agent CLIs (--agent random), avoiding the same backend on consecutive runs when another installed option exists

It normalizes the common controls that usually differ across these tools:

  • interactive vs non-interactive execution
  • initial prompt
  • working directory
  • permission/interaction mode
  • model class (fast or pro)
  • session resume
  • latest-session continue
  • non-interactive agent failover order

The repository also includes an optional provider-neutral MCP server for owner-scoped interactive session control across Cursor, OpenCode, Codex, and Claude. See [`mcp/README.md`](mcp/README.md) for official SDK/API setup, capability limits, and mocked tests.

By default, `agent-launch` starts agents in `auto` mode unless `AGENT_LAUNCH_MODE` or `--mode` overrides it.

Privacy-reduced transcript export

`agl export` detects supported local transcript stores and creates a ZIP for evaluating coding-agent proficiency. It keeps redacted user messages, short redacted assistant snippets, tool call names/counts, and day-level timestamps. It excludes raw transcripts, system/developer prompts, reasoning, tool arguments/results, working directories, account identifiers, and original session IDs.

Run