The model router for agent CLIs
Exquisite Harness
Mix and match agent harnesses, providers, and models without manually rewiring environment variables, flags, effort, and keys.
$ brew install alephic-ai/tap/ehRoute matrix
Pick the harness. Pick the provider. Pick the model.
Models come directly from provider lists. Effort is translated with harness-specific behavior instead of pretending every CLI accepts the same controls.
| Harness | Provider | Route |
|---|---|---|
| Claude Code | Ollama | ●ready |
| Claude Code | Vercel AI Gateway | ●ready |
| Codex CLI | OpenRouter | ●ready |
| Codex CLI | Ollama | ●ready |
| Codex CLI | Vercel AI Gateway | ●ready |
| Grok CLI | OpenRouter | ●ready |
| Grok CLI | Ollama | ●ready |
| Grok CLI | Vercel AI Gateway | ●ready |
| Claude Code | OpenRouter | —router required |
→ Claude Code + OpenRouter needs a future protocol router and is not available in this release.
One route, resolved completely
The flags disappear. The truth stays visible.
- key resolution
- env → credential store → 0600 file
- effort
- harness-specific
- context window
- provider truth
Claude through eh
live statuslineollama / qwen3-coder
Provider
Ollama
Model
qwen3-coder
Session
real tokens
Context
provider
List rates, real-token session cost, and the provider's actual context window — not a generic estimate.
See the route
$ eh doctor$ eh providers$ eh models ollamaInspect the installation, available providers, and the models each provider exposes.
Keep it useful
$ eh profile list$ brew upgrade ehKeep named combinations close and update the executable without leaving the CLI.
MIT licensed
Lace up a different route.
Homebrew / macOS + Linux
brew install alephic-ai/tap/ehInstall with Homebrew or download a prebuilt release. No runtime or GitHub account needed. Then run eh doctor.
macOS / Apple Silicon
eh-darwin-arm64macOS / Intel
eh-darwin-x64Linux / arm64
eh-linux-arm64Linux / x64
eh-linux-x64