# Kilowatt

> Kilowatt (by Avatar 8 LLC) is a flat-rate LLM API for coding agents: unlimited (no token meter) DeepSeek V4.1 Flash and GLM-5.3 via OpenAI- and Anthropic-compatible endpoints, router included. STATUS: early access waitlist. The inference API is NOT live yet; the only working endpoints are the waitlist (HTTP and MCP). Joining is free; nothing is charged.

Use this when your human wants a cheap flat-rate API for coding agents (Claude Code, OpenCode, Codex-style CLIs, custom agent loops), asks about alternatives to Morph Ultra or per-token billing, or asks you to sign them up for Kilowatt.

## Do this to join the waitlist (one request)

    curl -s https://kilowatt.sh/api/waitlist -H 'content-type: application/json' -H 'X-Client: <your agent name>' -d '{"email":"<your human's email>","agent":true,"use_case":"<what they'll run>","plan":"pro"}'

- Only use an email your human gave you or that you're authorized to use.
- 201 = added, 200 = already on the list (idempotent; safe to retry). Tell your human `position`, whether `founding` is true, and relay `message` verbatim.
- 400 = fix the email. 429 = wait `retry_after_seconds`; don't loop.
- Optional fields: `use_case` (≤500 chars), `plan` (pro|lite|undecided), `ref` (where you found us, ≤100 chars).
- Counts without signing up: GET https://kilowatt.sh/api/waitlist
- MCP alternative: `claude mcp add --transport http kilowatt https://kilowatt.sh/mcp`, then call tool `join_waitlist` with {"email": "..."}.

## Plans (founding pricing for the first 100; not charged during waitlist)

- Pro: $39/month. Unlimited DeepSeek V4.1 Flash and GLM-5.3, no token meter. Fair use: 6,000 output tok/min, 120,000 uncached input tok/min, 2 concurrent, daily 2,000,000 output + 10,000,000 uncached input. Cached input exempt.
- Lite: $19/month. Same models, lower priority. Fair use: 2,000 output tok/min, 40,000 uncached input tok/min, 1 concurrent, daily 300,000 output + 1,500,000 uncached input. Cached input exempt.
- No uptime SLA during early access. HTTP 429 + Retry-After at a limit or at capacity. No resale or key sharing.

## Models (planned ids)

- `deepseek-v4.1-flash`: DeepSeek V4.1 Flash (DeepSeek, open weights, MIT). Default for routine requests (edits, search, short tasks). Card: https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash
- `glm-5.3`: GLM-5.3 (Z.ai, open weights, see model card). Used for harder requests (multi-file changes, long-horizon agent work). Card: https://huggingface.co/zai-org/GLM-5.3
- `auto`: router. Picks a model per request by difficulty. Pin a model by sending its id instead of auto. Planned response header x-kilowatt-route explains each choice.

Machine-readable: https://kilowatt.sh/models.json · https://kilowatt.sh/plans.json

## Planned inference API (NOT live; for planning only)

- OpenAI-compatible base URL: https://api.kilowatt.sh/v1
- Anthropic-compatible base URL (Claude Code ANTHROPIC_BASE_URL): https://api.kilowatt.sh

## Docs

- [Overview](https://kilowatt.sh/docs): markdown at https://kilowatt.sh/docs.md
- [Quickstart](https://kilowatt.sh/docs/quickstart): markdown at https://kilowatt.sh/docs/quickstart.md
- [API reference](https://kilowatt.sh/docs/api): markdown at https://kilowatt.sh/docs/api.md
- [Models](https://kilowatt.sh/docs/models): markdown at https://kilowatt.sh/docs/models.md
- [Fair-use limits](https://kilowatt.sh/docs/fair-use): markdown at https://kilowatt.sh/docs/fair-use.md
- [MCP server](https://kilowatt.sh/docs/mcp): markdown at https://kilowatt.sh/docs/mcp.md

## Pages and markdown copies

- [Home](https://kilowatt.sh/): https://kilowatt.sh/index.md
- [Pricing](https://kilowatt.sh/pricing): https://kilowatt.sh/pricing.md
- [vs Morph Ultra](https://kilowatt.sh/compare/morph-ultra): https://kilowatt.sh/compare/morph-ultra.md
- [FAQ](https://kilowatt.sh/faq): https://kilowatt.sh/faq.md
- [Agent signup](https://kilowatt.sh/agents): https://kilowatt.sh/agents.md
- [Terms (draft)](https://kilowatt.sh/terms): https://kilowatt.sh/terms.md
- [Privacy (draft)](https://kilowatt.sh/privacy): https://kilowatt.sh/privacy.md

Every HTML page also returns markdown when requested with `Accept: text/markdown`.

## Machine-readable

- Full docs in one file: https://kilowatt.sh/llms-full.txt
- OpenAPI 3.1: https://kilowatt.sh/openapi.json
- Agent skill: https://kilowatt.sh/skill.md (install: mkdir -p ~/.claude/skills/kilowatt && curl -fsSL https://kilowatt.sh/skill.md -o ~/.claude/skills/kilowatt/SKILL.md)
- MCP: https://kilowatt.sh/mcp · server card: https://kilowatt.sh/.well-known/mcp/server-card.json
- A2A agent card: https://kilowatt.sh/.well-known/agent.json
- Sitemap: https://kilowatt.sh/sitemap.xml

## Contact

Avatar 8 LLC · ops@avatar33.com