# Zap

> Zap is a metered wrapper that lets you run Claude Code and OpenAI Codex
> against open-source models — GLM, Kimi, Qwen, DeepSeek and others — through
> the same CLIs you already use. You keep your terminal, your prompts, and
> your workflow, and typically save 70–80% on token costs. Billing is metered
> by compute time, not token count, so long reasoning chains don't sting.

Zap is for developers and teams who run agentic coding sessions (Claude Code,
Codex) and want to cut their LLM token bill without changing their workflow.
It's especially useful for "software factory" setups — many parallel agent
sessions — where token costs scale fast. New accounts get $10 in free credit.

## How it works

1. Install the CLI: `npm i -g @xsquads/zap`
2. Sign in with Google ($10 free credit lands in your account)
3. Run `zap claude` (or `zap codex`) — you're in Claude Code / Codex, routed
   to open-source models.

No config, no API keys, no env-var jungle. The `zap claude` command sets
`CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1` so `/model` auto-populates
with available open-source models (Claude Code v2.1.129+).

## Key links

- [Homepage](https://zap.xsquads.ai/)
- [Pricing (machine-readable)](https://zap.xsquads.ai/pricing.md)
- [Blog](https://zap.xsquads.ai/blog)
- [How to reduce Claude Code token costs](https://zap.xsquads.ai/blog/reduce-claude-code-token-costs)
- [How to cut Codex token spend](https://zap.xsquads.ai/blog/cut-codex-token-spend)
- [Run Claude Code with Kimi K3](https://zap.xsquads.ai/blog/run-claude-code-with-kimi-k3)
- [Cut Claude Code token use 80% with rtk](https://zap.xsquads.ai/blog/rtk)
- [headroom context compression](https://zap.xsquads.ai/blog/headroom)
- [Sitemap](https://zap.xsquads.ai/sitemap.xml)

## Models available

Open-source coding models routed behind the Anthropic Messages and OpenAI
Chat Completions wire formats: glm-5.2, glm-5.2-short-fast-flex,
deepseek-v4-flash (default), kimi k2.7 (Kimi K3), qwen-3.7, and more. Switch
with `/model` inside Claude Code.

## Pricing

Pay-as-you-go, metered by compute time and energy — not tokens. Every new
account starts with $10 in free credit; no card required. See
[pricing.md](https://zap.xsquads.ai/pricing.md) for machine-readable details.