The universal language for AI agents
75 lines teach any LLM to speak it. 10.41x compression. 100% parse validity. Zero dependencies.
Whitepaper · Rosetta · Documentation · Experiments · Website
What is AXL?
AI agents can't talk to each other. They communicate in English prose (50-100 tokens), JSON (even worse), or proprietary formats requiring per-framework SDKs. In a 100-agent network, this produces 22.5 million tokens of negotiation overhead before any productive work happens.
AXL eliminates this. One URL (@axlprotocol.org/rosetta) teaches any agent the complete language on first contact. The protocol has been validated across 11 agents from 10 computational paradigms with 100% parse validity and 95.8% LLM comprehension across four major architectures.
Install
pip install axl-core
Quick Start
Parse a packet:
from axl import parse
packet = parse("π:5:0xSIG:0.001|S:OPS.4|@api.example.com|ERR500|latency|4500ms|500ms|ALERT|URG|LOG")
print(packet.domain) # "OPS"
print(packet.body.fields) # ['@api.example.com', 'ERR500', 'latency', '4500ms', '500ms', 'ALERT']
print(packet.flags) # ['URG', 'LOG']
Emit a packet:
from axl import emit, Packet, Preamble, Body, PaymentProof
packet = Packet(
preamble=Preamble(payment=PaymentProof(agent_id="5", signature="0xSIG", gas=0.001)),
body=Body(domain="OPS", tier=4, fields=["@api.example.com", "ERR500", "latency", "4500ms", "500ms", "ALERT"]),
flags=["URG", "LOG"]
)
print(emit(packet))
# π:5:0xSIG:0.001|S:OPS.4|@api.example.com|ERR500|latency|4500ms|500ms|ALERT|URG|LOG
CLI:
axl parse "π:5:0xSIG:0.001|S:SIG.3|BTC|69200|↓|RSI|.64|SIG"
axl validate "π:5:0xSIG:0.001|S:SIG.3|BTC|69200|↓|RSI|.64|SIG"
axl translate --to english "π:5:0xSIG:0.001|S:SIG.3|BTC|69200|↓|RSI|.64|SIG"
Domains
| Domain | Purpose | Example |
|---|---|---|
OPS |
Infrastructure & operations | Server down, latency spike, disk full |
SEC |
Security & threat detection | Theft detected, balance discrepancy |
DEV |
Development & code lifecycle | PR merged, build failed, review needed |
RES |
Research & analysis | Cross-domain correlation, market analysis |
SIG |
Signal broadcast | BTC falling, RSI divergence |
COMM |
Communication & routing | Request, ACK, status update |
TRD |
Trading & economic action | Short BTC, 2x leverage, 2% max risk |
PAY |
Payment transfer | Pay agent $0.02 USDC for crawl task |
FUND |
Funding request | Request $5 for infrastructure costs |
REG |
Registration & identity | New agent joins the network |
The Stack
AXL is the missing language layer in the agent internet:
| Layer | Protocol | Who |
|---|---|---|
| Payment | x402 | Coinbase / Cloudflare |
| Tool Calling | MCP | Anthropic |
| Discovery | A2A | |
| Social | Moltbook | Meta |
| Language | AXL | AXLPROTOCOL INC |
Proven
Two live experiments. Real agents. Real packets. Real results.
| Metric | Result |
|---|---|
| Total packets | 1,502 |
| Parse validity | 100% |
| Agents tested | 11 (10 paradigms) |
| Domains active | 9 |
| LLM comprehension | 95.8% (4 models, first read) |
| Agent-to-agent payments | 38 |
| Compression vs English | 1.3-3x per message, 71x network |
Architecture
- Parser:
parse(string)-> Packet dataclass - Emitter:
emit(Packet)-> AXL string (round-trip proven) - Validator: Schema check, type check, tier range
- Translator: AXL <-> English <-> JSON
- CLI:
axl parse,axl emit,axl validate,axl translate - Zero runtime dependencies
v3 Support
Full v3 parser, emitter, validator, and translator included. Auto-detects v1 vs v3 format. All v1 code preserved for backward compatibility.
from axl import parse_v3, emit_v3, validate_v3, v3_to_json, v3_to_english
pkt = parse_v3("ID:MKT-01|OBS.99|$BTC|^67420|^fund:-0.02%+^OI:12.4B|NOW")
print(pkt.operation) # Operation.OBS
print(pkt.confidence) # 99
print(pkt.subject_value) # BTC
print(emit_v3(pkt)) # ID:MKT-01|OBS.99|$BTC|^67420|^fund:-0.02%+^OI:12.4B|NOW
print(v3_to_english(pkt)) # MKT-01 observes $BTC at ^67420 with 99% confidence.
print(v3_to_json(pkt)) # {"v":"3","id":"MKT-01","op":"OBS","cc":99,...}
print(validate_v3(pkt)) # [] (no errors)
Links
- Whitepaper - 14 sections, 5 appendices, every number measured
- The Rosetta v3 - 75 lines, the compressed kernel
- Evolution - From 300 lines to 75, the full narrative
- Battleground v2 Results - Full experiment data
- Terminal Recording - Watch the agents run
Contributing
See CONTRIBUTING.md for guidelines.
License
Apache 2.0 - AXLPROTOCOL INC · Diego Carranza · 2026
Metadata
Release files for axl-core 0.7.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| axl_core-0.7.0.tar.gz | 37.7 kB | Without provenance | Details for axl_core-0.7.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| axl_core-0.7.0-py3-none-any.whl | Python 3 | none | any | Details for axl_core-0.7.0-py3-none-any.whl |
Total release size: 72.1 kB
Release files / axl_core-0.7.0.tar.gz
| Download URL | axl_core-0.7.0.tar.gz |
|---|---|
| Size | 37.7 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
88d86d030588f7647012fdcca14b56a87bfeadca8e58e0d73a975dfd0042799b
|
|
BLAKE2b-256 checksum How to use checksums |
aec662d5ec4cd958a5010c3bef504f8637e854a9b2eecedd4cf096e0adfc057b
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.12.3
|
Release files / axl_core-0.7.0-py3-none-any.whl
| Download URL | axl_core-0.7.0-py3-none-any.whl |
|---|---|
| Size | 34.4 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
8dca9677058b7a742194b2c00080f86323c841d4168c640bb04e98d33aefdce3
|
|
BLAKE2b-256 checksum How to use checksums |
13094e043dcda07dd10eedc9f7baeffae9acb5f96d1a22f5f9992d649f75daab
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.12.3
|
Security for axl-core 0.7.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.