Update, Nov 9, 2026: CodeRouter's pricing has changed since this post. It is now pay as you go from a prepaid balance with no monthly fee, with smart routing or Direct mode (you name the exact model) on the same account, or Team per developer. The monthly plans and the separate Direct plan described below are no longer sold; accounts already on them keep their terms. The rest of the post is as published. Current terms: pricing.
Checked 2026-11-04 on LiteRouter's pricing page, docs, model list, terms and privacy policy. Prices and limits change, so check them before you buy. CodeRouter is our product, and TokenWatt, which ran the benchmark quoted below, is a measurement project from the same team.
TL;DR — LiteRouter is an OpenAI-compatible API aggregator, run by Eqlyx, LLC (Delaware), that sells daily request allowances rather than tokens: a free Basic plan, paid plans from $3 to $17 a month, and high-volume plans from $30 to $336 a month. Its pricing page lists Claude Sonnet from the $13 Elite plan and Claude Opus on the $17 Ultimate plan. Bigger requests use more of the allowance, and its premium and metered variants trim long chats to your "chat memory" setting. It fits chat apps and light, short-prompt use of many models for a fixed monthly price. For coding agents that send long, mostly cached contexts all day, billing by the token at list price is usually the better fit: CodeRouter also picks the model per request and keeps sessions on one model so the prompt cache stays warm, and OpenRouter lets you pick models yourself.
What LiteRouter is
- One API for many models. An OpenAI-compatible base URL,
https://api.literouter.com/v1, plus the Anthropic Messages endpoint (/v1/messages) for Claude Code. Its pricing page says 100+ models and its docs 400+; on 2026-11-04 the model list showed 1,733 model IDs, counting variants such as:metered,:full-contextand:payg. - Tools. Its docs have guides for Claude Code, Codex CLI, Cursor, Continue.dev, the OpenAI SDKs, Kilo Code, OpenCode, OpenClaw and Hermes Agent, and for chat and role-play apps such as SillyTavern, Janitor AI and Risu AI.
- Who runs it. Eqlyx, LLC, a Delaware company, per its terms (last updated March 2, 2026).
- Data. Its terms say it "does not store conversation content or message history"; it keeps operational data such as model, generation settings, provider error messages and token counts. Its privacy policy says requests are routed through third-party model providers.
LiteRouter pricing (checked 2026-11-04)
| Plan | Price | Premium requests a day | Concurrent requests | Model access (pricing page) |
|---|---|---|---|---|
| Basic | Free | Up to 50 | 1 | Free models (daily limit per model), 7 s cooldown |
| Standard | $3 a month | Up to 800 | 1 | |
| Plus | $6 a month | Up to 1,600 | 1 | Adds GPT 5 and Gemini 2.5 Pro |
| Pro | $10 a month | Up to 2,800 | 5 | Adds Gemini 3 Pro and Claude Haiku |
| Elite | $13 a month | Up to 3,700 | 10 | Adds Claude Sonnet |
| Ultimate | $17 a month | Up to 5,000 | 15 | Adds Claude Opus — "all models unlocked" |
| Paramount to Cosmos | $30–$336 a month | 9,000–108,000 | Top-plan features with higher limits |
- The pricing page shows each paid price next to a higher crossed-out one (Standard: $3, crossed out $4), so these may be promotional prices.
- Free models have no daily cap on paid plans, but the pricing FAQ notes a rate limit of 100 requests an hour.
- The model list labels some cheaper Claude variants with lower plans than the pricing page does (for example a
-cheapSonnet variant marked Plus), so check the exact model ID you need.
How a request is counted
The daily numbers above assume a "standard" request: model cost 1 and under 15,000 tokens. In practice:
- Regular variants: cost = model cost × an optimization cost that is 1 for the first 15,000 tokens and rises by 1 for every additional 5,000 (20,000 tokens → 2, 30,000 → 4). Claude, Gemini and GLM models have a higher model cost than cheap models.
- Full-context variants (
:full-context): one flat model cost per request, whatever the length. - Metered variants (
:metered,:metered:full-context): the model's price per 5,000 tokens, from the first token. - Context: premium and metered variants follow your chat memory setting — LiteRouter trims long chats to fit it — while
:full-contextvariants keep the model's full window. - Daily credits reset at 00:00 GMT+7 and don't roll over. Beyond them you can buy permanent credits: $1 buys 5,000, top-ups from $3 to $500, by card through Stripe or in USDT/USDC with a 5% discount.
:paygvariants bill per token (input, output and cached input priced separately) from those credits.
A worked example, using its published numbers. On 2026-11-04 the model list showed claude-sonnet-5.5:metered:full-context at a cost of 20.1 per 5,000 tokens. A request carrying 50,000 tokens would then use 10 × 20.1 = 201 premium credits, so Elite's 3,700 daily credits cover about 18 such requests; after that, each costs about 201 permanent credits, or $0.04. For scale, CodeRouter's own engineers made a median of 163 requests per developer per active day (75th percentile 378). Run your own request sizes through LiteRouter's cost calculator before you pick a plan.
Using LiteRouter with Claude Code and Cursor
Its Claude Code guide sets three variables:
export ANTHROPIC_BASE_URL="https://api.literouter.com"
export ANTHROPIC_AUTH_TOKEN="your_literouter_api_key"
export ANTHROPIC_MODEL="claude-sonnet-4.6:metered" # their example; any Claude model ID from the list
Per its docs, the Messages endpoint accepts only Claude models, and Claude models need a paid plan that includes them. For Cursor, its docs say Claude models are not supported: Cursor sends any model whose name contains "claude" straight to Anthropic, ignoring the custom base URL, so you use a non-Claude model there.
Two things to weigh for coding agents:
- Prompt caching. Claude Code resends a long, mostly unchanged prefix on every turn. Token billing with cache prices makes those repeats cheap; metered and flat variants are priced by request size instead (
:paygvariants do price cached input separately). - Concurrency. Basic, Standard and Plus allow one request at a time. Coding agents can send several at once — when they run subagents, for example.
LiteRouter vs CodeRouter
| LiteRouter | CodeRouter | |
|---|---|---|
| Billing unit | Requests (credits) a day on a monthly plan; :payg variants per token |
Dollars at the provider's list price |
| Plans | Free; $3–$17 a month; $30–$336 for high volume | Free trial ($5 of usage over 7 days, no card); Starter $19, Solo $29, Pro $99 a month with $15 / $25 / $85 of usage included, then list +5% up to a $40 / $60 / $200 cap; Direct: pay as you go at list +5% |
| Choosing a model | You pick the exact model ID and variant | model: auto — routing picks Claude Haiku 5.5, Kimi K2.6, Claude Sonnet 5.5 or Claude Opus 5.5 per request and moves up a tier only when tests or errors show a failure (Direct: you pick the exact model) |
| Prompt cache | Metered and flat variants priced by size | Cached input billed at the cache price; sessions stay on one model so the cache stays warm |
| Claude Code | Claude models only, paid plan needed | ANTHROPIC_BASE_URL=https://www.coderouter.io, ANTHROPIC_AUTH_TOKEN=cr_…, ANTHROPIC_MODEL=auto |
| Cursor | Non-Claude models only | OpenAI-compatible base URL https://www.coderouter.io/api/v1 |
| Teams | A key per project, revocable | Team: $19 per developer a month plus usage at list +5%; one key per engineer; budgets per member, project and company; email at 80%; downgrade or pause at 100% |
| Request content | Says it doesn't store conversation content | Stores a trimmed excerpt of each request for up to 180 days to improve routing; Team admins can turn that off; never sold |
How CodeRouter's routing performs: in TokenWatt's 176-task coding benchmark, it solved 96–100% of tasks for 56–70% less than sending every request to Opus, across runs (methodology and limits).
Which one fits
- LiteRouter fits chat and role-play apps, trying many models cheaply, and short prompts, when a fixed monthly price matters more than per-request cost and daily caps and cooldowns are acceptable.
- CodeRouter fits developers and teams running Claude Code, Cursor, Aider or Cline in long agent sessions who want the model chosen per request, billing at list price with prompt caching, and budgets per developer. For Claude Code specifically, see Claude Code Router: open-source vs hosted.
- OpenRouter fits picking from many models yourself at list price per token, with a 5.5% fee on credit purchases (checked 2026-11-04; comparison).
- LiteLLM fits teams that want to self-host a gateway with their own provider keys (comparison).
Before you buy any API aggregator
- Ask which upstream serves your model. LiteRouter's model-naming guide defines an
-officialtag for models that "run on the provider's own API"; on 2026-11-04 only eight model IDs, all DeepSeek, carried it. Our router safety checklist covers why this matters. - Estimate long requests, not just request counts, if your tool sends big contexts.
- Check the refund window. LiteRouter's terms say 3 days after purchase; its pricing FAQ says 7.
- Read the data policy for what is logged and for how long.
FAQ
What is LiteRouter?
An OpenAI-compatible API aggregator run by Eqlyx, LLC (Delaware) that gives one key and base URL (https://api.literouter.com/v1) for many models, billed as daily request allowances on monthly plans, with a free Basic plan (checked 2026-11-04).
How much does LiteRouter cost?
Basic is free. Standard is $3 a month (800 premium requests a day), Plus $6 (1,600), Pro $10 (2,800), Elite $13 (3,700, adds Claude Sonnet) and Ultimate $17 (5,000, adds Claude Opus); high-volume plans run $30–$336 a month. Requests over 15,000 tokens, and pricier models, count as more than one request.
Does LiteRouter work with Claude Code?
Yes, through its Anthropic Messages endpoint: set ANTHROPIC_BASE_URL=https://api.literouter.com, your LiteRouter key as ANTHROPIC_AUTH_TOKEN, and a Claude model ID as ANTHROPIC_MODEL. Only Claude models work there, and they need a paid plan that includes them.
What is a good LiteRouter alternative for coding agents?
If you want the model picked for you per request, CodeRouter routes across Claude Haiku 5.5, Kimi K2.6, Claude Sonnet 5.5 and Claude Opus 5.5 and bills at list price with prompt caching. If you want to choose models yourself, OpenRouter; to self-host, LiteLLM. More options in our AgentRouter alternatives guide.