Why Router9

The Problem

Running an AI harness — an autonomous agent or a coding CLI like Claude Code or OpenCode — means stitching together model providers, search APIs, and observability tools. Each piece has its own billing model, its own dashboard, and its own failure modes. The result is fragile infrastructure and unpredictable costs.

Common pain points:

• Per-token billing makes it impossible to budget for a busy harness
• No isolation between harnesses sharing the same API key
• No built-in audit trail for compliance and debugging
• Advanced capabilities (voice, vision, image generation) require separate SDKs and providers
• A light month is money wasted — unused budget evaporates instead of carrying forward

The Router9 Approach

Router9 is the Token Plan for AI Harness: a subscription-based gateway for any harness, agent or coding CLI. One plan, one predictable monthly fee, and as many API keys as you need.

One Token Plan, Many Keys

Each account has a single Token Plan with its own credit quota, rate limits, and audit trail. Issue a separate API key per harness or environment (up to 5) — revoke any one independently without affecting the others.

Every Model, Best Value

Router9 fronts multiple providers behind a single endpoint and routes each request for the best performance-per-dollar — GPT, Claude, Gemini, DeepSeek and hundreds more. A flat subscription buys cost-effective access to all of them.

Predictable Pricing — with Rollover

No per-token billing. No surprise costs. Each plan tier has a fixed monthly fee, a generous credit quota, and clear rate limits. And on paid plans you keep unused credits into the next month — 60% on Pro and 80% on Max, accumulating up to a hard cap of 2× your monthly quota — so quiet months build a buffer for busy ones.

All Skills, Every Tier

Every plan tier has access to every Skill: speech recognition, text-to-speech, image understanding, and image generation — with video generation and more on the roadmap. Skills are metered by your plan's shared credit allowance, so nothing is locked behind a paywall.

Drop-in Compatible

Router9 exposes an OpenAI-compatible API. Change your base_url and API key -- that is it. Works with Claude Code, OpenCode, Cursor, any OpenAI SDK, and any framework that supports the OpenAI chat completions format.

# Before
export OPENAI_BASE_URL=https://api.openai.com/v1

# After
export OPENAI_BASE_URL=https://api.router9.com/v1
export OPENAI_API_KEY=sk-r9k-your-key-here

Built-in Skills

No extra SDKs. Install the Router9 Skill package (or wire it up via MCP) and your harness gains ears (ASR), a voice (TTS), eyes (vision), and creativity (image generation) — with video generation and more on the way.

Deep Audit Tracking

Every request is logged with model, tokens, latency, cost, and status. Export CSV for compliance. Filter by API key, time range, or status. Retention scales with your tier.

Who is Router9 For?

• Solo developers who want to run an AI agent or coding CLI without managing infrastructure
• Teams who need to monitor and audit multiple harnesses in production
• Enterprises who require strict data isolation, predictable costs, and compliance-grade audit trails

Get Started

Create your first Token Plan in under a minute:

1. Sign up for a free account
2. Create a Token Plan from the dashboard
3. Copy your API key and set it as OPENAI_API_KEY
4. Point your harness at https://api.router9.com/v1

That is it. Your harness now routes through Router9.