Products

Tools for compressing, deliberating, and connecting AI agents.

RESEARCH PREVIEW

AXL Rosetta v4.0.1 Kernel-Router

/rosetta/v4/

Kernel-router architecture with pluggable Rosetta modules (financial, construction, prose fallback). Shared canonical form layer, artifact-driven gating, drift detection. 217/217 tests passing. Frozen at v4.0.2-r6 (commit 51e75de).

Qualified successor verdict: on domain-backed content, v4 replaces v3.1 (corpus #1 +15.02 recall / +14.54 precision; corpus #2 +36.64 / +43.96). On prose fallback, v4 is recall-favored but precision-down; v3.1 remains preferred for precision-sensitive narrative use cases until the gap closes.

PACKET SYNTAX PKT[VER|CLS|SUB|TAG|ARG1|ARG2|META]

Seven pipe-delimited fields. v3 parsers extract fields legally without interpreting v4 extension semantics.
Research stage. v3.1 remains the productized stable release.
v4 Hub → Read Kernel Spec →
LIVE

AXL Compress

compress.axlprotocol.org

Paste English prose. Get a compressed version optimized for LLM context windows. 2-3x on single messages, up to 10x on multi-field structured data. Works with ChatGPT, Claude, Gemini, Grok - any LLM.

Compression is domain-dependent. Structured data with multiple fields compresses most (5-10x). Prose paragraphs compress modestly (2-3x). The LLM reads the Rosetta header, becomes fluent instantly, and processes your compressed content.

ENGLISH - 67 tokens "Magnesium acts as a cofactor for over 300 enzyme systems that regulate diverse biochemical reactions in the body, including protein synthesis, muscle and nerve function, blood glucose control, and blood pressure regulation."
AXL - 28 tokens π:BIO-01|OBS.95|#Mg_cofactor|^300_enzyme_systems|$protein_synth+$nerve_func+$glucose_ctrl+$BP_reg|NOW

2.4x on this excerpt. Compression varies by content structure and domain.
Free & open source (Apache 2.0)
OPEN SOURCE

AXL Swarm

pip install axl-swarm

Multi-agent deliberation engine. 12 experts debate your question across 12 rounds. They disagree, synthesize, change their minds, and converge on a consensus prediction. Same engine that produced 10.41x compression on medical diagnosis.

ONE COMMAND pip install axl-swarm
axl-swarm run seed.md --compare

8 pre-built seeds: finance, medicine, military intelligence, science, philosophy, geopolitics, legal, personal career decisions. Write your own seed for any domain.

Works with any LLM: Claude, GPT, Llama, Qwen, Gemini - via litellm. Run local with Ollama for $0.

Free & open source (Apache 2.0)
GitHub →
BETA

Sophon

sophon.axlprotocol.org

Intelligence dashboard for agent swarms. Watch agents think in real-time. See consensus form. Track belief changes. Identify influence chains. Scrub through deliberations frame by frame.

Web monitor (any browser) + GPU-rendered 3D visualization (Blender + RTX 5090). Every node is an agent. Every edge is an interaction. Click any node to see everything that agent said.

Enterprise solutions
GitHub →
ROADMAP

AXL Bridge

bridge.axlprotocol.org

Make any AI agent talk to any other AI agent. LangChain ↔ CrewAI ↔ AutoGen ↔ MetaGPT ↔ ElizaOS. One protocol. Zero integration code. Each agent gets an AXL endpoint. The Rosetta handles translation.

Enterprise solutions
ROADMAP

AXL API

api.axlprotocol.org

Compression-as-a-service. Sits between your application and your LLM provider. Compresses system prompts, RAG context, and conversation history before each API call. Your $10,000/day token bill becomes $2,500/day.

Free & open source (Apache 2.0)