I-Lang v3.0 — released · MIT License

Protocols
for the AI era.

Eastsoft Inc. is a Canadian research company building open communication protocols for human ↔ AI and AI ↔ AI systems. We don’t compete with foundation models. We make them speak the same language.

See our portfolio Read our story →
✓
Verified on 7 LLMs
∞
Zero install · pure text
protocol.ilang
# I-Lang v3.0 — AI-native protocol
# Read all .md, merge, output:

[LIST:@LOCAL|mch=*.md]
  => [Π:READ]
  => [Σ]
  => [Ω]

# Plain English equivalent: 23 tokens
# I-Lang equivalent:  8 tokens
Token savings, verified by tiktoken −65%
2008
Founded
7
LLM platforms verified
88
Protocol verbs
−65%
Token savings
MIT
Open source
Portfolio

One protocol. A growing toolchain.

Each product ships I-Lang to a different surface — IDE, browser, terminal, agent. Built and maintained inside Eastsoft; released open-source.

⬢ Flagship · Research

I-Lang Protocol

The native language of artificial intelligence. 88 verbs, two syntaxes, zero install. Reduces semantic loss between human intent and machine execution. AI-to-AI, AI internal, human-to-AI.

READWRITFMTMAPFILT SORTΣ MERGEθ XLATΩ OUT + 79 more
ilang.ai →
[ ]
{ }
Plugin · Claude Code

AutoCode

You say it, AutoCode ships it. 47 skills detect what you're building, compress every instruction.

GitHub →
◉
URL Proxy · Vision

AI See

Give any model eyes. Prefix any URL with i.ilang.ai/ — paste into chat, the model reads.

i.ilang.ai →
::
Behavioral profile

Imprint

AI learns how you work, not what you did. One portable file across every agent. 312 tokens.

GitHub →
⌘
ClawHub · Skill bundle

OpenClaw Skills

Instruction-only skills published on ClawHub. Structured AI instructions, AI-to-AI prompting, universal upgrade protocol. All security-scan green.

GitHub →
∅
IDE Extension

ZeroCode

From zero to live website. AI decides everything; you say yes or no. Built on the I-Lang spec, distributed via Trae and VS Code.

GitHub →
Why this matters

A standard layer between humans and AI.

HTTP didn’t invent the web. It gave the web a shared grammar. I-Lang is built from symbols already inside every LLM’s training data. No SDK, no runtime, no dialect.

−65%
verified token savings

Measured with OpenAI tiktoken (cl100k_base) across hundreds of real-world prompts. Less spend, faster turns, fewer rewrites.

7 of 7
major LLMs

ChatGPT, Claude, Gemini, DeepSeek, Kimi, Qwen, GLM. The protocol activates with a single paste. No fine-tune, no vendor lock-in.

Nature
acknowledged

Our paper on AI hallucination was acknowledged by Nature for technical quality. Available on ResearchGate, SSRN, ChinaXiv, and arXiv.

Research & Press

Peer-reviewed. Published. Open.

"We're not building another model. We're building the layer that lets every model agree on what was said."
— Eastsoft research note · 2026
arXiv preprint"Reducing semantic loss in human-AI communication"
Read →
ResearchGateEastsoft research mirror
Visit →
SSRN / ChinaXivCross-region archives
Visit →
Hugging FaceProtocol artifacts & datasets
Open →
GitHub Organizationilang-ai · MIT licensed
Open →
Since 2008

Eighteen years of quiet building.

Eastsoft started as a software services and distribution shop. Today we are an AI‑native research company. We don’t raise rounds. What we look for is partnership, and the right rooms.

Read our full story →
2008
Eastsoft incorporated as a software services & distribution company.
2014
Crossed into international markets. Began publishing third-party software globally.
2020
Internal R&D pivots toward AI‑native tooling and protocol research.
2024
First public draft of I-Lang. NASDAQ:JZ shareholding established.
2026
I-Lang v3.0. Research arm spins up at research.ilang.ai.

Talk to Eastsoft.

We're not raising. We're looking for partners, distribution, and rooms where standards get written. If you have any of those — say hello.

[email protected] View on GitHub →