Toknix is a unified AI model gateway engineered for high-concurrency developer workflows, autonomous agent harnesses, and enterprise infrastructure. With a single secret API key, developers can route traffic across 40 leading foundation models from Anthropic, OpenAI, Google, xAI, DeepSeek, Zhipu AI, and Moonshot AI.
Overview
Instead of maintaining separate billing relationships, API keys, SDK dependencies, and rate limit monitors across multiple AI providers, Toknix provides a single, low-latency edge proxy. The gateway handles request translation, automatic prompt caching, failover routing, and prepaid token metering.
Zero Email & Pure Crypto Deposits
Request Formats & Base URLs
Toknix accepts both OpenAI and Anthropic standard wire formats. Existing client libraries, official SDKs, and developer tools work out-of-the-box by updating just the Base URL and API Key:
| Protocol Format | Endpoint | Base URL to Configure |
|---|---|---|
| OpenAI Chat Completions | POST /v1/chat/completions | https://api.toknix.in/v1 |
| OpenAI Responses API | POST /v1/responses | https://api.toknix.in/v1 |
| Anthropic Messages API | POST /v1/messages | https://api.toknix.in |
| Model Catalog | GET /v1/models | https://api.toknix.in/v1 |
| Account & Balance | GET /v1/balance | https://api.toknix.in/v1 |
Every public model is cross-compatible across all formats. You can invoke Claude models through OpenAI Chat Completions, or invoke DeepSeek and GLM models through Anthropic Messages.
Core Infrastructure
- 40 Frontier & High-Speed Models: Unified access to Gemini 4, GPT-6, Claude Opus/Sonnet, Grok 4.7, DeepSeek V4.1, GLM 5.3, and Kimi K3.
- Uncensored & Abliterated Weights: Specialized abliterated models for vulnerability testing, red-team security audits, and unrestricted creative synthesis.
- Automatic Prompt Caching: Frequently recurring context prefixes receive automatic cache hits with discounts up to 90% (down to $0.003/1M tokens on GLM 5.3 Flash).
- Prepaid Crypto Billing: Add balance starting from $5 with automatic on-chain credit and bonus tiers up to 16%. Failed requests with 5xx errors are never charged.
- Subkeys & Team Quotas: Provision independently funded subkeys with spend caps for team members, CI/CD pipelines, or autonomous agent loops.
Start Building
Follow our quickstart guides to start making calls or configure your favourite coding agent: