# NeverMetered

> Flat-rate, OpenAI- and Anthropic-compatible inference for coding agents. No token meter and no 5-hour or weekly limits; your plan sets how many agents run at once.

## Product
- [Home](https://nevermetered.com/)
- [Pricing](https://nevermetered.com/pricing)
- [How it runs](https://nevermetered.com/how-it-runs)
- [Status](https://nevermetered.com/status)
- [About](https://nevermetered.com/about)
- [Sign up](https://nevermetered.com/signup)
- [Compare](https://nevermetered.com/compare)
- [Qwen3.8-27B](https://nevermetered.com/models/qwen3-8-27b)

## Docs
- [Documentation](https://nevermetered.com/docs)
- [Integration guides](https://nevermetered.com/docs/integrations)

## Integrations
- [Claude Code](https://nevermetered.com/docs/integrations/claude-code): Anthropic's terminal coding agent, pointed at your own base URL and model.
- [Codex CLI](https://nevermetered.com/docs/integrations/codex-cli): OpenAI's open-source terminal agent, using a custom Responses API provider.
- [Copilot CLI](https://nevermetered.com/docs/integrations/copilot-cli): GitHub's terminal coding agent, using Copilot CLI's official Bring Your Own Key (BYOK) support.
- [OpenCode](https://nevermetered.com/docs/integrations/opencode): Open-source terminal agent with an OpenAI-compatible provider.
- [Hermes Agent](https://nevermetered.com/docs/integrations/hermes): Nous Research's open-source terminal agent, using a custom OpenAI-compatible endpoint.
- [OpenClaw](https://nevermetered.com/docs/integrations/openclaw): The open-source, always-on personal AI agent, using a custom OpenAI-compatible provider.
- [DeepSeek Harness](https://nevermetered.com/docs/integrations/deepseek-harness): DeepSeek's open-source, plugin-based coding agent (the `dsh` command), pointed at a custom endpoint.
- [Pi](https://nevermetered.com/docs/integrations/pi): The minimal, extensible terminal coding agent from pi.dev, with a custom OpenAI-compatible provider.
- [Oh My Pi](https://nevermetered.com/docs/integrations/oh-my-pi): A batteries-included fork of Pi (the `omp` command), configured with a custom provider in models.yml.
- [Aider](https://nevermetered.com/docs/integrations/aider): AI pair programming in your terminal, through Aider's OpenAI-compatible support.
- [Cline](https://nevermetered.com/docs/integrations/cline): Autonomous coding agent for VS Code. The same steps work for Roo Code and Kilo Code.
- [Cursor](https://nevermetered.com/docs/integrations/cursor): Use the gateway for Cursor chat through its OpenAI base URL override.
- [Continue](https://nevermetered.com/docs/integrations/continue): Open-source assistant for VS Code and JetBrains, configured in config.yaml.
- [OpenAI SDK](https://nevermetered.com/docs/integrations/openai-sdk): The official Python and TypeScript clients, with Chat Completions or the Responses API.
- [Anthropic SDK](https://nevermetered.com/docs/integrations/anthropic-sdk): The official Python and TypeScript clients for the Messages API.
- [Vercel AI SDK](https://nevermetered.com/docs/integrations/vercel-ai-sdk): TypeScript toolkit for AI apps, via the OpenAI-compatible provider.
- [LangChain](https://nevermetered.com/docs/integrations/langchain): LangChain and LangGraph in Python through ChatOpenAI.
- [LiteLLM](https://nevermetered.com/docs/integrations/litellm): Call the gateway from the LiteLLM SDK or put it behind a LiteLLM proxy.
- [curl](https://nevermetered.com/docs/integrations/curl): Raw HTTP requests, for testing or for languages without an SDK.

## Guides
- [Guides](https://nevermetered.com/guides)
- [Claude Code usage limits: 5-hour and weekly caps explained](https://nevermetered.com/guides/claude-usage-limits): How Claude Code's 5-hour session and weekly limits actually work in 2026, why routine work burns through them fast, and where a flat-rate lane fits.
- [Claude Code pricing: every plan explained](https://nevermetered.com/guides/claude-code-pricing): Real Free, Pro, Max, Team, Enterprise, and API pricing for Claude Code in 2026, how the usage limits work, and when a flat-rate lane makes sense.
- [Codex & ChatGPT Usage Limits Explained](https://nevermetered.com/guides/codex-usage-limits): How OpenAI's Codex CLI and ChatGPT Plus/Pro usage limits work: the 5-hour window, the weekly cap, what changed in 2026, and what to send elsewhere.
- [Cursor usage limits explained](https://nevermetered.com/guides/cursor-usage-limits): How Cursor's usage pools, rate limits, and Auto pricing work, and where a flat-rate second lane fits for the routine agent work Cursor doesn't cover.
- [Best local LLM for coding in 2026 (and when hosted beats local)](https://nevermetered.com/guides/best-local-llm-for-coding): Real 2026 hardware costs, VRAM tiers, and speeds for running a coding LLM locally, plus when a flat-rate hosted option beats buying a GPU.
- [Codex CLI vs Claude Code: which should you use in 2026?](https://nevermetered.com/guides/codex-vs-claude-code): A neutral, fact-checked comparison of Codex CLI and Claude Code: how each works, current pricing, benchmark scores, and who each tool actually fits.
- [Claude Code vs Cursor: terminal agent or IDE assistant?](https://nevermetered.com/guides/claude-code-vs-cursor): Claude Code runs as a terminal-first coding agent; Cursor is a VS Code fork built around inline AI. How each works, what it costs, and who it fits in 2026.
- [OpenAI API pricing in 2026: every model, explained](https://nevermetered.com/guides/openai-api-pricing): OpenAI's per-token API pricing for 2026: GPT-6 Astra, Sol and Luna, GPT-5.6 Terra, GPT-5.4 and o-series, plus caching, batching and long-context surcharges.
- [Anthropic Claude API pricing in 2026: every model, explained](https://nevermetered.com/guides/claude-api-pricing): Anthropic's per-token Claude API pricing across Opus 5.5, Opus 5, Sonnet, Haiku and Fable, plus prompt caching, Batch API discounts and rate limits by tier.
- [Claude Code subagents explained (and why they burn your usage limit fast)](https://nevermetered.com/guides/claude-code-subagents): What Claude Code subagents are, how they isolate context, and why running several in parallel can drain a 5-hour or weekly usage window fast.
- [OpenRouter alternative for coding agents: flat-rate vs per-token](https://nevermetered.com/guides/openrouter-alternative): How OpenRouter's per-token billing works in 2026, why it's hard to predict for coding agents, and where a flat-rate, priced-by-streams alternative fits.
- [GitHub Copilot usage limits explained (and how to stop hitting them)](https://nevermetered.com/guides/github-copilot-usage-limits): How GitHub Copilot's premium requests and AI Credits work today, and how Copilot CLI's BYOK feature routes around the limit.
- [OpenCode vs Claude Code: open-source vs official terminal agent](https://nevermetered.com/guides/opencode-vs-claude-code): OpenCode is a free, open-source terminal agent that works with any model. Claude Code is Anthropic's own agent, tied to a Claude plan. Here's how they differ.
- [Windsurf vs Cursor: which AI IDE should you use in 2026?](https://nevermetered.com/guides/windsurf-vs-cursor): Windsurf (now Devin Desktop) vs Cursor for 2026: pricing tiers, agent models, IDE support, and BYOK limits compared side by side, no fluff.
- [Cursor vs GitHub Copilot: which fits your workflow?](https://nevermetered.com/guides/cursor-vs-copilot): A neutral comparison of Cursor's AI-native editor and GitHub Copilot's extension-plus-CLI approach: pricing, usage limits, and workflow differences.
- [What is a context window? LLM context explained for coding agents](https://nevermetered.com/guides/context-window-explained): What an LLM context window is, why coding agents fill it faster than chat, and what compaction and context rot mean for picking a model or plan.
- [Cline vs Cursor: VS Code extension or dedicated AI IDE?](https://nevermetered.com/guides/cline-vs-cursor): Cline vs Cursor, compared fairly: BYOK VS Code extension vs. bundled-model AI IDE. 2026 pricing, model flexibility, and who each one actually fits.
- [Aider vs Claude Code: which terminal AI pair programmer wins?](https://nevermetered.com/guides/aider-vs-claude-code): Aider's diff-based git commits vs Claude Code's autonomous tool loop: how pricing, model choice, and workflow compare, and who each actually fits.
- [What developers actually say about AI coding agent limits](https://nevermetered.com/guides/ai-coding-limits-developer-sentiment): We read through the real Reddit threads about Claude Code, Codex and Cursor limits. Real quotes, real numbers, on what's actually driving the complaints.
- [What running an AI coding agent actually costs, per day](https://nevermetered.com/guides/ai-coding-agent-daily-cost): Real per-token API rates for Qwen3.8-27B and cheap open models across a dozen providers, and what a 24/7 coding agent actually costs at market rates.
- [How fast is Qwen3.8-27B really? What local benchmarks actually show](https://nevermetered.com/guides/qwen3-8-27b-local-speed-benchmarks): Real local Qwen3.8-27B speed tests range from 34 to 127 tok/s on the same GPU class, depending entirely on tuning. Here's why, and where hosted sits.

## Legal
- [Terms of Service](https://nevermetered.com/terms)
- [Privacy Policy](https://nevermetered.com/privacy)
- [Refund Policy](https://nevermetered.com/refunds)
- [Acceptable Use Policy](https://nevermetered.com/acceptable-use)