Introduction to Tokligence Gateway

Tokligence Gateway is a high-performance AI Gateway for coding agents and enterprise token management.

Vision

Three pillars for the AI-native era:

Your AI agents handle sensitive code, secrets, and business data. Tokligence protects them:

PII Prompt Firewall - Real-time detection and redaction of sensitive data across 100+ languages
API Key Protection - Detects 30+ provider keys (OpenAI, AWS, GitHub, Stripe, etc.) before they leak to LLM providers
Multiple Modes - Monitor, enforce, or redact based on your compliance needs
Seamless Integration - Works with Codex CLI, Claude Code, and any OpenAI/Anthropic-compatible agent

Think of Tokligence as a buffer for your AI throughput:

Peak Hours - Buy tokens from the marketplace when internal LLM capacity is maxed out
Off-Peak - Sell your unused LLM throughput to earn revenue
Elastic Scaling - No need to over-provision; scale with actual demand

Not just another gateway - the foundation for AI token economics:

Unified Access - OpenAI, Anthropic, Gemini with bidirectional protocol translation
Token Ledger - Built-in accounting and audit trail for every token consumed or sold
Open Source - Apache 2.0, self-hosted, no vendor lock-in
High Performance - 9.6x faster than LiteLLM with 75% less infrastructure

Both OpenAI and Anthropic APIs running simultaneously with zero adapter overhead:

OpenAI-Compatible API: /v1/chat/completions, /v1/responses, /v1/models, /v1/embeddings
Anthropic Native API: /anthropic/v1/messages, /anthropic/v1/messages/count_tokens
Gemini Native API: /v1beta/models/* endpoints

Based on benchmarks against LiteLLM (v0.3.4):

Metric	LiteLLM (4 instances)	Tokligence (1 instance)	Improvement
Throughput	1,170 RPS	11,227 RPS	9.6x faster
P50 Latency	100 ms	49.66 ms	2x faster
P99 Latency	240 ms	93.81 ms	2.6x faster
Infrastructure	4 instances	1 instance	75% reduction