itsjustanks

Agent Link — Development skill for Claude Code

Development community

Run multiple Claude Code & Codex accounts at once — parallel, hot-switched, or auto-routed.

How to install Agent Link

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

What Agent Link does

Run multiple Claude Code & Codex accounts at once — parallel, hot-switched, or auto-routed. One slot per account, zero credential copying. macOS + Linux.

Alternatives in Development

CORS — Scan an endpoint for CORS misconfiguration — arbitrary-origin reflection, null-origin trust, credential exposu 4.5k ★
Claude Account — Switch between isolated Claude Code accounts on Linux and macOS without logging in again 73 ★
Equip Silo Provider — Add a provider (claude, codex, grok, gemini, …) to an already-onboarded account silo with farm equip, completi 67 ★

README

agent-link

**Run several Claude Code and Codex accounts from one machine — and see what your agents actually build.**

[![License](https://img.shields.io/badge/license-MIT-blue)](LICENSE) ![Platform](https://img.shields.io/badge/macOS%20%C2%B7%20Linux-informational) ![Dependencies](https://img.shields.io/badge/deps-bash%20%2B%20python3-lightgrey) ![Paseo](https://img.shields.io/badge/Paseo-plugin%20included-8A63D2)

  CLAUDE                              plain `claude` → primary
    ● you@work.com                   primary        in rotation · 12 launches
    ● you@home.com                   logged in      in rotation · 9 launches · last 4m ago
    ● you@side.com                   logged in      spend limit reached — routing skips it
    dynamic routing: ready (~/.agent-link/bin/claude-auto)

Three problems this solves

**You can only use one account at a time.** If you have more than one Claude or Codex account, you hit a usage limit on one while the others sit idle. Account-switcher tools swap credentials in and out, which logs running sessions out mid-task. agent-link instead gives each account its own config directory — `CLAUDE_CONFIG_DIR` for Claude Code, `CODEX_HOME` for Codex — so every account is live at the same time. **N accounts, N rate limits.** It never reads, copies or backs up a token, so nothing decays and no running agent is ever signed out.

**A usage limit kills your afternoon.** As an account climbs toward its limit, agent-link drains new work off it; when it hits the wall it is parked until the provider's own reset time; and a conversation that died mid-task continues on a healthy account with a plain `--resume` — or resumes itself, if it was a [Paseo](https://paseo.sh) agent. Claude reports through its statusline and hooks, Codex through its own rollout files. You stop noticing limits.

**Your agents write things you never see.** Dashboards, reports and diagrams land in a worktree you would have to go