OmniRouteby Cheaper Inference

免费开源 AI 网关

永不停止编程。

Route across 352 providers through one OpenAI-compatible endpoint, with automatic fallback.

Start with Cheaper InferenceSelf-host OmniRoute (opens in a new tab)
  • 352 providers
  • 90+ free tiers
  • 15–95% eligible tokens saved
OpenAI-compatible Free and open source59.6k stars (opens in a new tab)

OmniRoute × Cheaper Inference

Combine OmniRoute’s automatic fallback and developer tools with Cheaper Inference’s cost-ranked marketplace. Every eligible request starts with the cheapest available provider route.

6 models
Direct list prices compared with Cheaper Inference prices by model
ModelInput priceOutput priceSavings
gpt-5.6-lunaOpenAI · text$0.20$0.08/ 1M tokens$1.20$0.48/ 1M tokens60% off
gpt-5.6-solOpenAI · text$2.00$1.00/ 1M tokens$10.00$5.00/ 1M tokens50% off
glm-5.3Z.ai · text$1.40$0.77/ 1M tokens$4.40$2.42/ 1M tokens45% off
deepseek-v4-flashDeepSeek · text$0.083$0.05/ 1M tokens$0.18$0.11/ 1M tokens40% off
deepseek-v4-flash-0731DeepSeek · text$0.067$0.04/ 1M tokens$0.18$0.11/ 1M tokens40% off
glm-5.3-flashZ.ai · text$0.071$0.05/ 1M tokens$0.24$0.17/ 1M tokens30% off
Direct list price → Cheaper Inference price · View live catalog (opens in a new tab)Pricing checked Aug 31, 2026, 12:00 AM UTC. Rates can change when upstream pricing or available routes change.

承诺

单一端点。永不停止构建。

永不触碰限制

毫秒内在 339 个提供商间自动切换。配额用尽？下一个提供商接管. 零停机。

最多节省 95% 的 Token

A 12-engine pipeline reduces eligible context by 15–95%. The documented tool-heavy example reduced it by ~89%.

$0 起步

90+ free tiers and 56 recurring or keyless free-forever providers. No credit card needed for the keyless quick start.

35 supported integrations

16+ 个编程助手. Claude Code、Codex、Cursor、Cline、Copilot、Antigravity. 通过一个配置。

单一端点

OpenAI ↔ Claude ↔ Gemini ↔ Responses API 翻译。将任何工具指向 /v1，即可正常工作。

生产级

熔断器、TLS 隐身、MCP（105 个工具）、A2A、记忆、护栏、评估. 由 25,000+ 个测试支撑。

提供商目录

All behind one endpoint.

A catalog spanning chat, media, search, local runtimes, cloud agents and system integrations.

352 providers

14 Search12 Audio-only2 Upstream proxy3 Cloud agent1 System
25OAuthClaude Code, Codex, Cursor, GitHub Copilot, Kimi Code…
233API keyOpenAI, Anthropic, Groq, NVIDIA, Cerebras, Mistral…
35Web cookieChatGPT Web, Claude Web, Gemini Web, Grok Web…
14LocalOllama, LM Studio, vLLM, llama.cpp…
13No authenticationOpenCode Free, Felo, Pollinations and other keyless routes…
Explore the provider reference (opens in a new tab)

60 秒内启动运行

三条命令，零配置。

Install the package, launch the gateway, then point any OpenAI-compatible client at port 20128.

Provider catalog
352
Routing strategies
19
Local endpoint
:20128/v1

Step 1: Install

Install the `omniroute` npm package globally.

omniroute: install
$ npm install -g omnirouteadded 1 package

Step 2: Run

Launch the API and dashboard together on port 20128.

omniroute: run
$ omniroute▸ dashboard ✓ http://localhost:20128/dashboard▸ api ...... ✓ serving on :20128

Step 3: Connect your tool

Use `http://localhost:20128/v1`; the keyless `auto` route works on a fresh install.

omniroute: connect
$ curl localhost:20128/v1/models✓ models listed

开箱即用

一切内置。无需插件。

路由、容错、免费配额消耗、压缩和智能体协议。单一二进制文件，零插件拼接。

智能组合

Set a model to `auto` or build your own combo with 19 strategies and tiered fallback.

Use a ready-made mode or combine 19 strategies around the goal that matters.

记忆

持久对话记忆. FTS5 关键词 + Qdrant 向量召回，内置。

3-layer resilience

在正确的层级失败。而非整体崩溃。每个提供商有熔断器，每个连接有冷却，每个模型有锁定。其他密钥和模型持续服务。

免费配额

Pool-deduped accounting across 90+ free tiers.

每月约 1.51B 免费 Token。

压缩

12 composable engines for tool output, context and response style.

Reduce eligible tokens by 15–95%.

MCP 服务器

A built-in server exposing 110 gateway tools across 33 scopes.

110 tools · 33 scopes · stdio / HTTP

A2A 协议

A JSON-RPC agent server with 6 skills.

6 skills · JSON-RPC 2.0 · Agent Card

CombosResilienceFree quotaCompressionMCPA2AMemoryProxy controlsGuardrailsCloud agentsEvalsSemantic cacheSkillsAnalytics

Client compatibility

Keep your existing coding tools.

Point an OpenAI-compatible CLI or editor at one local base URL; OmniRoute handles the providers behind it.

Claude Code
Codex CLI
Cursor
Cline
GitHub Copilot
Gemini CLI
OpenCode
Kilo Code
Factory Droid
Continue
Roo Code
Antigravity
Aider
OpenClaw
Goose
Grok Build
Windsurf
Devin CLI

http://localhost:20128/v1

request.sh
curl http://localhost:20128/v1/chat/completions \
  -H "Authorization: Bearer $OMNIROUTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"auto","messages":[{"role":"user","content":"Hello"}]}'

新

现已支持 VS Code Copilot Chat

安装 OmniCopilot 扩展，每个 OmniRoute 模型都会直接出现在你已经在用的模型选择器里。不需要新的侧边栏，不需要新的聊天界面。

Official Marketplace

Works in VS Code stable and Insiders. Install from the Marketplace or Extensions view.

Open VSX Registry

适用于不使用官方商店的 VS Code 分支。Cursor、Windsurf、VSCodium、Theia、code-server、Gitpod 等。

无需 Copilot 套餐

从 VS Code 1.122 起，提供商模型无需 GitHub 登录或 Copilot 订阅即可使用。在 OmniRoute 的免费模型上支持 agent mode、工具调用和视觉能力。

VS Code Marketplace (opens in a new tab)Open VSX Registry (opens in a new tab)OmniCopilot source (opens in a new tab)

日常真实痛点

为何选择 OmniRoute

每一个 AI 编程痛点都有内置解决方案。无需脚本，无需人工看护。

Without routing
With OmniRoute
Quota expires unused
Track and drain the allowance you already have
Rate limits stop you mid-task
4-tier auto-fallback in milliseconds
Tool output burns tokens
12-engine compression, 15–95% on eligible content
Expensive APIs add up
Cost-optimized routing can pick the cheapest eligible target
Each tool needs its own setup
One endpoint and one dashboard
A provider is unavailable from your network
Proxy scopes and client-fingerprint controls
One dead key breaks a provider
Breaker, cooldown and lockout isolate the failure
A team shares one subscription
Key pools and fair-share quotas
You need local control of routing data
Local-first storage with encrypted credentials
No spend or health visibility
Usage, quota, savings and latency analytics
1. Subscription2. API key3. Cheap4. Free

诚实对比

竞品对比

Capabilities checked against the linked project documentation on 2026-08-30.

CapabilityOmniRoute (opens in a new tab)9Router (opens in a new tab)LiteLLM (opens in a new tab)CLIProxyAPI (opens in a new tab)
Providers35240+100+8+ upstreams
Routing strategies193-tierretry / priorityround-robin / fill-first
Tiered fallback with UIYesYesManual policyMulti-account
Token compression12 engines · 15–95%RTK / Headroom / CavemanNo built-in pipelineHTTP compression

Public README and official documentation snapshot; verify linked sources before purchase or deployment decisions.

随处运行

9 supported ways to run OmniRoute.

同一套代码，九种运行方式。笔记本、服务器、手机或插件。选择安装命令，即可在 20128 端口提供服务。

npm

npm

npm install -g omnirouteOpen guide (opens in a new tab)

Docker

Docker

docker run … diegosouzapw/omnirouteOpen guide (opens in a new tab)

Desktop

桌面端

npm run electron:buildOpen guide (opens in a new tab)

ARM

ARM

native arm64

Android · Termux

Termux

pkg install nodejs && npx -y omnirouteOpen guide (opens in a new tab)

PWA

PWA

Add to Home ScreenOpen guide (opens in a new tab)

OpenCode plugin

OpenCode 插件

@omniroute/opencode-providerOpen guide (opens in a new tab)

VS Code

VS Code

Install OmniCopilotOpen guide (opens in a new tab)

From source

源码

npm install && npm run devOpen guide (opens in a new tab)

Routing

Choose an `auto` route.

Use a ready-made mode or combine 19 strategies around the goal that matters.

Model · fallback enabledautoBalanced default
OpenAI-compatible request
curl http://localhost:20128/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"auto","messages":[{"role":"user","content":"Write a TypeScript retry helper."}]}'

Routing alias auto selected. Objective: Balanced default. Fallback enabled.

Drain subscription quota

Use quota you already have before touching paid APIs.

priority · fill-first

Spread load

Balance traffic across keys and accounts to reduce rate-limit pressure.

weighted · round-robin · p2c · least-used · headroom

Cheapest first

Route to the lowest-cost eligible target.

cost-optimized

Context-aware

Keep long contexts on models that fit, then relay when needed.

context-relay · context-optimized · cache-optimized

Randomized

Distribute eligible traffic across targets.

random · strict-random

Adaptive scoring

Use 15-factor scoring, last-known-good and reset-aware windows.

reset-window · reset-aware · lkgp · auto · fusion · pipeline

免费 Token，诚实统计

每月约 1.51B 免费 Token。

455 catalog entries are deduplicated into 40 recurring pools; only published positive budgets feed the headline.

1.51B

Pool-deduped accounting keeps shared quotas from being counted more than once. No-cap providers stay visible without inflating the token total. Checked 2026-08-30.

View the calculation (opens in a new tab)
Recurring or keyless free-forever providers56
  • Mistral1B/month
  • LLM7150M/month
  • Nara150M/month
  • Gemini CLI · forever60M/month
  • Kiro50 credits/month
  • Qoder · foreverFree catalog
  • LongCat · forever50M/day
  • Cerebras1M/day
  • NVIDIA NIM~40 RPM
  • Pollinations · foreverNo key
  • Cloudflare Workers AI · foreverDaily neurons
  • OpenCode Free · foreverNo-auth catalog

停止浪费 Token

Reduce eligible tokens by 15–95%.

工具输出、重复上下文和臃肿的 JSON 在到达模型前就被压缩。语义不变，Token 仅为原来的一小部分。12 个可组合引擎在您掌控的流水线中依次叠加。

Before69 tokens
The function should iterate over the list of items, and for each item, it should check whether the value is greater than zero, and if so, add it to the running total.
After19 tokens
sum all list items where value > 0

Composable pipeline

  1. Session-Dedup
    Drop content already sent earlier in the session.
  2. CCR
    Archive large blocks and restore them on demand.
  3. Lite
    Apply conservative lossless cleanup.
  4. RTK
    Filter and deduplicate command and tool output.
  5. Responses Tool Output
    Compact Responses API tool payloads.
  6. Headroom
    Apply reversible tabular and structured-data compaction.
  7. Relevance
    Keep sentences most relevant to the latest query.
  8. Caveman
    Use rule-based prose condensation.
  9. Aggressive
    Apply stronger configured pruning.
  10. LLMLingua-2
    Use code-safe semantic pruning in an isolated worker.
  11. Ultra
    Combine high-savings stages with preservation guards.
  12. OmniGlyph
    Encode eligible repeated structures compactly.

Profiles

Published profile range~30%x-omniroute-compression: standard

Standard compression selected. Published range ~30%.

Controls and output styles

  • Named profiles and active selector
  • Visual Compression Studio
  • Anthropic Context Editing
  • Adaptive context-budget dial
  • Per-request `x-omniroute-compression` header
  • Offline eval harness
  • Token-threshold auto-trigger
  • Terse prose
  • Less code (YAGNI)
  • Terse CJK 文言

Control precedence: request header → combo → profile → adaptive policy → default.

3 层，而非 3 级

在正确的层级失败。而非整体崩溃。

一个坏密钥不应拖垮整个提供商。一个失效模型不应拖垮密钥。每次失败都被隔离在最小可能的范围内，其余一切照常服务。

model ⊂ connection ⊂ provider
Provider circuit breakerWhen a provider is unhealthy, combo routing skips it until a half-open probe succeeds.
Connection cooldownCool down one key or account while sibling connections keep serving.
Model lockoutLock one rate-limited or unavailable model while other models on the connection remain available.

Network controls

Network routing and local data controls.

Proxy scopes control network routing while local-first storage keeps operational data on your machine.

Three proxy scopes

Route globally, per provider or per individual connection.

  • Global · per-provider · per-connection
  • SOCKS5 and HTTP(S) proxies
  • Proxy marketplace integration

TLS and client fingerprint controls

Match supported provider clients down to selected transport characteristics.

  • JA3/JA4-capable transport via wreq-js
  • Provider-specific client profiles
  • IPv4 and IPv6 egress controls

Private and local-first

Keys, usage and history stay on the host you control.

  • SQLite on disk with AES-256-GCM encrypted credentials
  • No OmniRoute account required
  • Self-host on infrastructure you choose
RussiaChinaIranCubaTürkiye

一览无余

完整控制台，而非配置文件。

Manage providers, combos, analytics, health and costs at `localhost:20128/dashboard`.

localhost:20128/dashboardExample data
352providers connected
90+free tiers available
198healthy in this example
  • Claude Code · OAuth · live
  • Cerebras · free · live
  • Groq · cooldown example
  • NVIDIA NIM · free · live

原生智能体支持

连接智能体。让它驱动 OmniRoute 本身。

OmniRoute 不只是代理。它支持智能体协议。将其接入 MCP 客户端、A2A 网络或云端编程智能体，它就成为智能体可以操控的工具。

MCP 服务器

110 tools across 33 scopes over stdio and HTTP transports.

110 tools · 33 scopes · stdio / HTTP

claude mcp add omniroute --type http --url http://localhost:20128/api/mcp/stream

A2A 服务器

6 skills for smart routing, quota, discovery, cost analysis and health reporting over JSON-RPC 2.0.

6 skills · JSON-RPC 2.0 · Agent Card

云端智能体

通过统一界面驱动 Codex、Cursor、Devin 和 Jules。创建任务、审批计划并流式传输结果，全部经由 OmniRoute 路由。

Codex · Cursor · Devin · Jules

超越路由

不止是代理。更是平台。

嵌入式服务

将 Bifrost（Go 中继）、9Router 或 CLIProxy 作为托管 sidecar 运行。或运行完整集群配置。全部直接从控制台操作。

你的知识，融入上下文

Notion 与 Obsidian 作为 MCP 工具接入。你的笔记与知识库成为任何模型的一等上下文。

内置护栏

每条路由都有提示注入防护，可选启用的 PII 脱敏，以及安全内容过滤归一化。有防御，无锁定。

游戏化

连续打卡、成就与实时节省。让你的用量看得见。

Guides

Setup guides in three languages

Walkthroughs cover installation, free-provider setup and connecting an IDE.

开放构建

一个快速发展的活跃社区

OmniRoute 持续更新. 由数百名贡献者推动，保持领先。

  • 8,274forks
  • 583contributors
  • 280releases shipped

Snapshot from 2026-09-01

Discord

Chat, support and roadmap

(opens in a new tab)

Telegram

News and announcements

(opens in a new tab)

WhatsApp 🌍

Global community

(opens in a new tab)

WhatsApp 🇧🇷

Comunidade Brasil

(opens in a new tab)
Meet the contributors (opens in a new tab)

OmniRoute by Cheaper Inference

Choose how you run OmniRoute.

Self-host OmniRoute, or start with Cheaper Inference for a hosted cost-ranked gateway.

Start with Cheaper Inference Self-host OmniRoute (opens in a new tab)