Use Claude Code with Manifest
Claude Code reads its API config from ~/.claude/settings.json
on every run. Drop the Manifest block in there and your Claude Code sessions quietly
route through the cheapest capable model. Still Claude when it matters, something
cheaper when it doesn't.
— models routed through Manifest over the last 30 days.
Why use Claude Code with Manifest
Claude Sonnet and Opus are top-tier reasoners but pricing scales fast on long agentic sessions. Manifest detects which prompts genuinely need Claude and routes the rest to cheaper models, without ever leaving the Claude Code experience.
- Slash Claude API costs by 40-70% with no code changes
- Persistent setup via settings.json, no shell rc edits
- Fall back to other providers when Anthropic rate limits hit
- Track every Claude request, model used, and cost in one dashboard
- Use Claude for hard prompts, cheap models for boilerplate
- Self-host the router on your own infra, no vendor lock-in
Setup in 60 seconds
Add a single block to ~/.claude/settings.json and
every claude run picks up the new endpoint.
No flag, no env var, no restart.
Add this block to ~/.claude/settings.json and
every claude run will route through Manifest.
{
"env": {
"ANTHROPIC_BASE_URL": "https://api.manifest.build",
"ANTHROPIC_AUTH_TOKEN": "mnfst_YOUR_KEY"
}
}
https://api.manifest.build
mnfst_YOUR_KEY
Claude Code tokens processed on Manifest
Models in use
# Model |
Auth type | Tokens / last 30 days |
|---|---|---|