Open source · Apache-2.0
The open-source next-best-action decisioning platform
KaireonAI is an Apache-2.0, self-hostable platform that decides the single best offer, message, or action for each customer in real time — with visual decision pipelines, nine scoring engines, and full control of your own data.
What open source changes about decisioning
Decisioning that affects real customers should be something you can understand and control. KaireonAI is open source under Apache 2.0 and self-hostable, so the stages, the models, and the decision traces are all meant to be visible rather than hidden behind a service. You run it on your own infrastructure and keep customer data inside your own environment — no black box, no per-decision metering, and no vendor holding your data.
It ships as a single application that serves both the UI and the decision APIs, backed by PostgreSQL. That means one thing to deploy on Linux, Docker, or Kubernetes, and one place to inspect exactly how every recommendation was chosen.
A decision pipeline you can see
Every request flows through a visual pipeline you build without code. Each stage narrows or scores the set of candidate actions, and the last stage picks a winner:
Inventory
The full set of candidate offers, messages, or actions for this touchpoint.
Eligibility
Fit filters remove anything the customer is not eligible for.
Contact policy
Frequency caps, suppression windows, and cooldowns protect the relationship.
Match scoring
A model scores each surviving candidate for this specific customer.
Ranking
Multi-objective ranking orders the candidates and selects the best.
Nine scoring engines, built in
The scoring stage is a choice, not one algorithm. Assign any of nine engines to a decision — each scores inside the pipeline, with no external inference hop.
Scorecard
Transparent, weighted rules. Ideal when you need an explainable, auditable score with no training data.
Bayesian / Naive Bayes
Fast, probabilistic scoring that learns continuously from outcomes as they arrive.
Logistic Regression
A dependable, calibrated baseline for binary response prediction.
Gradient Boosted Trees
Captures nonlinear feature interactions when you have richer data and want accuracy.
Thompson Bandit
Bayesian exploration that balances learning about new actions against exploiting known winners.
Epsilon-Greedy
A simple, robust bandit that reserves a slice of traffic for exploration.
Neural Collaborative Filtering
Learns latent customer and offer affinities for recommendation-style problems.
Online Learner
Updates incrementally on every outcome, so the model tracks shifting behavior without a batch retrain.
External Endpoint
Call your own model over an API when you already have one you trust.
Everything around the decision
A decision engine is more than a model. KaireonAI ships the whole stack open source: contact policies and suppression to prevent fatigue, A/B experiments with holdout groups and uplift measurement, 50+ data connectors for object stores, warehouses, streams and CRMs, real-time dashboards, and 160+ MCP tools with an AI assistant so you can build and maintain decisioning config in natural language. Security is on by default: per-tenant Row-Level Security, AES-256-GCM encryption at rest, and RBAC with an audit trail.
Frequently asked questions
Is KaireonAI really open source?
Yes. KaireonAI is licensed under Apache 2.0 and is self-hostable. It ships as a single Next.js application that serves both the UI and the decision APIs, so you can run the entire next-best-action platform on your own infrastructure.
What is next-best-action decisioning?
Next-best-action (NBA) decisioning is the discipline of choosing, in real time and per person, the single best offer, message, or action to take right now. It folds propensity, eligibility, contact policy, and business value into one ordered decision rather than a raw list of likely clicks.
Can I self-host the whole platform?
Yes. Because it is a single application backed by PostgreSQL, you deploy it on your own Linux, Docker, or Kubernetes environment and keep customer data inside your own environment. Nothing about the pipeline, models, or decision traces is hidden behind a hosted service.
How does an open-source platform score decisions?
Scoring is a choice, not a single algorithm. The platform ships nine scoring engines — from a transparent Scorecard to Gradient Boosted Trees, bandits, and an External Endpoint — that run inside the decision pipeline, per candidate, per request.
Keep reading: What is Next-Best-Action? · Platform features