ExLuna-rs/claude-autorouter
Auto-route Claude Code prompts to Haiku, Sonnet or Opus. Learns from escalations, saves tokens.
ARCHETYPE
Pragmatist
A balanced, no-drama setup: some rules, some tools, nothing extreme.
Copy this rig
# review before running: this installs third-party code $ npx degit ExLuna-rs/claude-autorouter/agents ./rig-claude-autorouter/agents $ npx degit ExLuna-rs/claude-autorouter/commands ./rig-claude-autorouter/commands $ npx degit ExLuna-rs/claude-autorouter/hooks ./rig-claude-autorouter/hooks
MCP servers are added to Claude Code at local scope; env vars are shown as YOUR_… placeholders — we never store values. Files are fetched with degit into a separate folder so you can review before merging.
$ curl -fsSL --create-dirs -o .claude/agents/architect.md https://raw.githubusercontent.com/ExLuna-rs/claude-autorouter/main/agents/architect.md $ curl -fsSL --create-dirs -o .claude/agents/builder.md https://raw.githubusercontent.com/ExLuna-rs/claude-autorouter/main/agents/builder.md $ curl -fsSL --create-dirs -o .claude/agents/quick.md https://raw.githubusercontent.com/ExLuna-rs/claude-autorouter/main/agents/quick.md
This rig commits no guardrails. Here is the community baseline instead — the deny/ask rules most often found across all 16.2k rigs:
{
"permissions": {
"deny": [
"Read(./.env)",
"Read(./.env.*)",
"Read(**/.env)",
"Read(~/.ssh/**)",
"Bash(git push --force:*)",
"Read(**/*.pem)",
"Read(.env)",
"Read(~/.aws/**)",
"Read(**/*.key)",
"Read(**/.env.*)",
"Bash(git push --force*)",
"Bash(rm -rf:*)",
"Bash(rm -rf *)",
"Bash(sudo *)",
"Bash(sudo:*)",
"Bash(rm -rf /)",
"Bash(git push -f:*)",
"Read(.env.*)",
"Bash(git reset --hard*)",
"Bash(rm -rf /*)"
],
"ask": [
"Bash(git push:*)",
"Bash(git push *)",
"Bash(gh pr merge *)",
"Bash(git commit:*)",
"Bash(rm *)",
"Bash(gh pr merge:*)",
"Bash(gh pr create *)",
"Bash(git reset --hard:*)",
"Bash(npm publish:*)",
"Bash(git rebase *)"
]
}
} Subagents (3)
| architect model: opus | Top tier for hard tasks - design and architecture, choosing between approaches, hard or intermittent bugs, security audits, authentication and cryptography, performance and concurrency, migrations, re |
| builder model: sonnet | Standard implementation tier - build a feature, fix an identified bug, write or repair tests, localised refactor across one or a few files, scripts and configuration. Use when the router says tier=bui |
| quick model: haiku | Fast, cheap tier for simple tasks - locate or search code, read and summarise a file, answer a factual question about the repo, rename, fix a typo, format, add a comment or an import. Use when the rou |
Slash commands (3)
/router-learn/router-stats/statusline
Similar rigs
matthieurosset/claude-custom-workflow
Multi-agent orchestration workflow template for Claude Code — one lead agent, six specialist subagents, guardrails, hook-driven quality gates, and a self-improving process loop. Battle-tested on a production app, generic and ready to adopt.
Orchestrator 447 tok · · Q15
aravindhsampath/agentic-template
Cookie-cutter project template for an agentic development workflow: Claude Code architect + engineer subagents, Pi-based multi-model code review (Deepseek + Kimi), strict branch hygiene, declarative apply-standard script.
Pragmatist 2.5k tok · · Q22
lahma/claude-roslyn-lsp
Roslyn C# language server adapter for Claude Code and other agents (LSP + MCP), .NET 10 Native AOT
Pragmatist 36.5k tok · · Q8
yourconscience/tackroom
Dotfiles for your AI agents: one ~/.agents repo for skills, MCP servers, hooks, agent roles, and memory — synced natively to Claude Code, Codex, OMP/Pi, OpenCode, Droid, and Hermes. Commit-pinned and audited externals.
Orchestrator 1.4k tok · · Q8