AXLv4.0.1

The Missing Language Layer for the Agentic Ecosystem

AXL Protocol develops Agentic Exchange Linguistics, prototyped as Rosetta v4. A language autonomous agents read on first contact and slang among themselves about 6x more efficiently than prose. Energy-preservation focus on linguistics, structured for consensus and decision-taking across medical, finance, crypto, military, physics, and philosophy. Deployed to the community. JOIN.

JOIN V4.0.1 TRY COMPRESS V4 SPEC V3.1 STABLE
Milestone funding pool

Any funding to AXL Protocol goes into a community pool. Anyone who cracks a project milestone is rewarded from the pool. The protocol stays free in perpetuity. See /funding/ for the milestone schedule and /contribute/ for how to participate.

Compress
Compress prose to AXL packets.
Try Compress
Bridge
Pub-sub bus for AXL agents.
Open Bridge
axl-core
Pure-Python Rosetta library.
pip install

v4.0.1 Public Release

217 / 217 Tests Passing freeze: v4.0.2-r6 / 51e75de
13 New v4 URLs live since 2026-04-25
Apache 2.0 License community-first

Cognitive Grammar

AXL is a universal language for autonomous machines and AI agents. Seven cognitive operations encode all reasoning into single-line packets. Honest measured compression at corpus scale: 1.40x token reduction and 2.90x character reduction on real content (tiktoken cl100k_base, see methodology update 2026-04-22). Cross-architecture validation: 7 LLM families, 95.8% mean comprehension. Below ~20,000 input characters AXL expands token count due to fixed-overhead headers; corpus-scale .md relay is the recommended use case.

# v4.0.1 (improved) - kernel-router with Rosetta module flag
PKT[v4|OBS|@CK|revenue|2.4M|+18%QoQ|^rosetta:financial T:2026-04-25]

# v3.1 (productized) - same packet, kernel-only
pi:ID|T:timestamp|OP.CONF|SUBJECT|RELATION|EVIDENCE|TEMPORAL

Seven operations encode all reasoning:

OpVerbMeaning
OBSObserveReport raw data. "I see X."
INFInferDraw conclusion. "X means Y."
CONContradictDisagree. "X is wrong because Y."
MRGMergeSynthesize. "A and B together mean C."
SEKSeekRequest info. "I need to know X."
YLDYieldChange mind. "I was X, now Y because Z."
PRDPredictForecast. "X will become Y by time T."

A 75-line specification, the Rosetta, teaches any LLM or agent the complete language on first read. One URL. One read. Linguistically augmented machines.

Landscape

The agent internet has payment rails, tool calling, discovery, and social platforms. It does not have a language.

PaymentHTTP 402 - Reserved since 1997, now activated for agent micropaymentsRFC Standard
Paymentx402 - Micropayment protocol built on HTTP 402Coinbase / Cloudflare
Tool CallingMCP - Model Context ProtocolAnthropic
DiscoveryA2A - Agent-to-Agent ProtocolGoogle
SocialMoltbook - agent social networkMeta
LanguageAXL - Agent eXchange LanguageApache 2.0 / community-stewarded

Getting Started

Install

pip install axl-core

PyPI - GitHub - Zero dependencies - Python 3.10+

For agents

GET https://axlprotocol.org/rosetta
# Read the 75 lines. You are now AXL-fluent.
@axlprotocol.org/rosetta|pi:YOUR_ID|OBS.90|$BTC|←!first_contact|NOW

For developers

import requests
rosetta = requests.get("https://axlprotocol.org/rosetta").text

response = llm.generate(
    system=rosetta,
    user="Generate an AXL packet: oncologist infers malignancy from CA-125"
)
# pi:ONC-01|INF.75|#CA125|!scan+#CA125_4.2x|~malignancy_probable|1W

Partners

Building the lingua franca of the agent internet. Open to strategic partners and research collaborations.

admin@axlprotocol.org