Aroora AI Labs
Homepage
A · manifesto
B · flagship
Independent research lab · aroora.ai

Building the open agentic stack.

Aroora AI Labs ships open infrastructure for trustworthy agentic AI — a token-efficient format (ISON), property graphs for context (ISONGraph), a multi-agent protocol engine (MAPLE), deterministic context shaping (Contexel) and local multi-model inference (SnapLLM). Repository first, product second.

Explore ISON →
All repositories
Aroora mark
Aroora mark
{{ statRepos }}
public repos
5
organisations
7
languages
11
publications
“AI systems should be verifiable, not just impressive.”

The stack

index of all repositories →
{{ r.name }} {{ r.badge }}
{{ r.tagline }}
{{ r.lang }} {{ r.org }} {{ r.stars }}
01 / Verifiable by design

Canonical serialization, byte-identical context shaping, replayable agent runs — every layer of the stack can be checked, diffed and reproduced.

02 / Token economics matter

Context windows are the scarce resource. ISON cuts payloads 30–70% versus JSON; ISONGraph fits 3× more graph into the same window.

03 / Open by default

MIT-licensed specs, reference implementations and reproducible benchmarks — private work becomes public the moment it is stable.
Flagship · v1.0.4 · PyPI npm crates.io

ISON

A minimal, token-efficient data format optimized for LLMs and agentic AI workflows.

Familiar tabular patterns models have seen billions of times. 30–70% fewer tokens than JSON, lossless round-trip, and byte-identical canonical output across all seven implementations.

Explore the repo →
pip install ison-py{{ copyLabel }}
MIT licence7 implementations860+ tests
json → ison 
{{ jsonSample }}
{{ isonSample }}
87 tokens34 tokens · −61%
And the rest of the lab
{{ r.name }}
{{ r.tagline }}
{{ r.lang }}

Repository index

Everything the lab maintains, across every organisation. Private repositories are marked and can be hidden — future repos slot into the same template.

{{ t.label }}
Show private
{{ resultCount }}

{{ g.org }}

{{ g.count }}
{{ r.name }} {{ r.badge }}
{{ r.tagline }}
{{ t }}
{{ r.lang }} {{ r.updated }} {{ r.stars }} {{ r.rel }}
Repo template
A · editorial
B · docs shell
{{ repoOrg }} / {{ repoName }}
{{ repoLive }}
private — opens when stable

{{ repoTitle }}

{{ repoHero }}

${{ repoInstall }}{{ copyLabel }}
{{ repoSiteBtn }} ↗ Source ↗
{{ f.k }}
{{ f.v }}

Playground

json → ison, right here in the page
input · json▶ convert
output · ison{{ playSaving }}
{{ playOut }}

Demo

product walkthrough slot — drop a screenshot or GIF

Quickstart

{{ t.label }}
 
{{ tabCode }}

README

condensed from {{ repoOrg }}/{{ repoName }}@main

{{ repoName }}

{{ rmIntro }}

Why

  • {{ w }}
    •

Install

{{ rmInstall }}

Status

{{ rmStatus }}

Architecture

architecture diagram — drop it here

Maintainer

MV
Mahesh Vaikri
AI Architect @ Dell Technologies · Founder, Aroora AI Labs · 11 publications

{{ repoName }}

{{ repoHero }}

{{ repoSiteBtn }} ↗
{{ m }}
${{ repoInstall }}{{ copyLabel }} 
{{ jsonSample }}
{{ isonSample }}

Quickstart

{{ t.label }}
 
{{ tabCode }}

Architecture

architecture diagram

Maintainer

MV
Mahesh Vaikri
Founder, Aroora AI Labs · GitHub ↗

Lab notes

Releases, benchmark results and format decisions — written when something ships, not on a schedule.

{{ p.date }}
{{ p.kind }}

{{ p.title }}

{{ p.excerpt }}

read note →
← all notes
{{ postKind }}{{ postDate }}Aroora AI Labs

{{ postTitle }}

{{ b.text }}

 
{{ b.text }}
Next note
{{ nextTitle }}
→

In the lab

What is being built next, published early so people can tell us we are wrong — or ask to help. Nothing here is a commitment.

{{ i.stage }} {{ i.eta }}

{{ i.name }}

{{ i.desc }}

express interest →

Get in touch

Commercial licensing, research collaboration, sponsorship, or a question about a spec — pick the closest inbox.

Research & sponsorship
research@aroora.ai
MAPLE commercial licence
mahesh@mapleagent.org
Name
Work email
What is this about
{{ t.label }}
Message
Send message