The lingua franca of the agent internet.
Three billion autonomous AI agents are projected to be active by 2028. They negotiate contracts, monitor supply chains, diagnose patients, and coordinate logistics. They cannot talk to each other across frameworks. Every agent interaction begins with 2,210 tokens of negotiation overhead.
AXL Protocol eliminates this overhead. A 377-line specification teaches any LLM the complete language on first read. Agents communicate in compressed packets - 10x fewer tokens, 2x faster, identical output quality. The protocol is self-bootstrapping: every packet carries the specification URL, so the language propagates through communication itself.
Self-bootstrapping acquisition. Every AXL packet carries the Rosetta URL. Every LLM that reads a compressed document learns the protocol. Every user who compresses a prompt teaches AXL to every LLM they touch. The language spreads through communication itself - the students become professors. No other protocol has this property.
AXL Protocol is a registered Canadian company based in Vancouver, BC. The protocol was developed and validated over a 6-day intensive sprint (March 17-22, 2026) using a multi-server infrastructure including bare-metal GPU server with 32GB VRAM running DDR7 at max clock, 16 PCIe lanes fully saturated, 98GB DDR5 system memory producing optimized tokenization outputs with local LLM models running billions of parameters in inference, extended through millions of Anthropic API calls to Claude Sonnet 4.6 to scale the neural network of the agents, coordinated through multiple experiments, multiple pipelines, multiple instances and iterations over and over, to make sure they could speak.
We are raising a seed round from Canadian and US venture firms.