We build open-source tools that help AI systems do useful work with less unnecessary inference, clearer execution control, and verifiable outcomes.
Website · KORA · Research paper · X / Updates
- Featured projects — open-source tools you can inspect and try
- Research — the KORA paper, evidence, and boundaries
- Get started — an example you can run locally
- Connect — community and website
Build with the tools that fit your workload. Each project has its own setup, scope, and documentation.
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
Albert Heekwan Kim · arXiv:2610.35869 · Read the paper · PDF · Source code
The paper describes KORA's deterministic-first execution-control architecture: route structured or directly verifiable work before requesting model inference, validate outcomes, and escalate when required.
In a reproducible 100-task deterministic-heavy simulated benchmark, KORA-controlled execution avoided 80 of 100 simulated model invocations relative to a naive direct baseline. This is a bounded benchmark result, not evidence of production savings, real API-cost reduction, universal quality improvement, or broad workload representativeness.
Try a local KORA workload inspection example. The example requires Python 3.10+ and does not need an API key or model download.
git clone https://github.com/Krako-Labs/KORA.git
cd KORA
python3 -m venv .venv
source .venv/bin/activate
python3 -m pip install -e .
python3 -m kora doctor examples/kora_doctor/customer_support_workload.jsonFull setup and limitations · Standalone KORA Doctor






