Research · Papers

The FlowX.AI paper series.

Reliability, governance, measurement and memory for enterprise agents — each engineered and shown running in production, not hoped for. Read a summary, then download the full technical paper.

Browse the papers See the open models

engineered · measured · in production

VERA
Evidence-Graded ROI for Production Agents
Bringing counterfactual measurement, causal validation, and confidence-graded accounting to the business case for enterprise AI agents.
Read summary PDF ↓
ORNA
Agent Self-Adaptation in Production
Closing the loop between observability, validation, and node-level refinement for LLM agents in regulated industries.
Read summary PDF ↓
HALO
Zero Hallucination, by Construction
Why trustworthy enterprise AI is a property of the harness around the model, not of the model itself: grounding, multi-signal verification, calibrated abstention, and continuous oversight in the FlowX.AI Platform.
Read summary arXiv ↗ PDF ↓
GAVEL
Governance That Runs Where the Agent Runs
Why governing autonomous AI agents means enforcing policy and producing audit evidence at runtime, over the agent s own telemetry, rather than asserting compliance on paper.
Read summary PDF ↓
SIFT
A Classifier That Teaches Itself
A cost-tiered document classifier whose corpus grows from production traffic and whose autonomous retraining is made safe by a frozen evaluation gate.
Read summary arXiv ↗ PDF ↓
RAILS
The Graph Is Deterministic, Even When the Nodes Are Not
How Agent Builder makes enterprise AI agents reliable by compiling visual workflows into deterministic state machines, in which the stochastic model is one bounded, typed node among many.
Read summary PDF ↓
MNEMĒ
A Knowledge Graph Agents Can Write To, Safely
MNEME, the FlowX.AI Ontology Layer: a provenance-tracked structured-memory graph that agents read, validate against, and contribute to, without corrupting the institutional source of truth.
Read summary PDF ↓

Open tools

Open guardrails for LLM I/O.

Beyond the papers, border is FlowX.AI's open guardrail layer — input and output checks for LLM applications, each with signed, verifiable evidence. Small models you can run yourself.

border.flowx.ai

LLM input & output checks with signed evidence

Open detectors on Hugging Face — prompt-injection, groundedness, moderation, toxicity, bias, PII and more.

Visit border

The FlowX.AI paper series.

Evidence-Graded ROI for Production Agents

Agent Self-Adaptation in Production

Zero Hallucination, by Construction

Governance That Runs Where the Agent Runs

A Classifier That Teaches Itself

The Graph Is Deterministic, Even When the Nodes Are Not

A Knowledge Graph Agents Can Write To, Safely

Open guardrails for LLM I/O.

LLM input & output checks with signed evidence