Product
ProcessProtectReconComply
Use Cases
PSPs & Payment ProcessorsFintechs & Digital BanksFinancial InstitutionsEnterprise TreasuryTransit & Municipal
Developer
API-FirstMCP for Coding AgentsSelf-Hosting & DeployArchitecture & Stack
Built for AI
Meet FloKnowledge BaseInternal MCP ToolsContext-Aware UIDocument Intelligence About
Book a DemoTry Free →
AI · Woven Through the Platform

An Analyst Inside the Ledger

AI in LedgerFlow isn't a bolted-on chatbot. Flo reads your ledger through governed tools, every operator page carries its own context, your documentation answers questions directly, and document intelligence turns KYC paperwork into structured data — all behind the same permissions and audit trail as everything else.

Florence — Your In-Platform Analyst

Flo is a conversational assistant built into the operator console. Ask a question in plain language and Flo calls real, read-only tools against your ledger to answer it — then shows its working.

Ask in Plain Language
"What were yesterday's largest transactions?" "Which accounts are fraud-flagged?" Flo turns the question into the right tool calls — no SQL, no endpoint hunting.
A Tool-Calling Loop
Flo runs an agent loop over a registry of read tools, chaining several calls when a question needs them — and never invents data it didn't retrieve.
History & Starter Prompts
Conversations are saved to a History tab, and curated starter prompts get new users productive immediately — both managed from within the app.

Flo tells you exactly what it's doing — the chat sidebar shows a different status for each kind of work:

Flo is thinking… Flo is accessing data… Flo is observing… Flo is SIEM'ing…

Answers Grounded
in the Documentation

A retrieval-augmented knowledge layer lets Flo — and connected coding agents — answer product questions from LedgerFlow's own documentation, not from guesswork.

Semantic Search
Documentation is embedded into pgvector and queried by meaning, so a question finds the right passage even when the wording doesn't match.
Grounded Responses
Retrieved passages are handed to the model as context, so answers stay tied to what the platform actually does — with the source material behind them.
Public & Internal Corpora
The corpus is split so external coding agents only ever see content marked agent-visible, while operators get the fuller internal knowledge base.

One Console for
"Is It Healthy? Is It Under Attack?"

Beyond the ledger, Flo can reach operational tools through the internal MCP server — so an operator asks, in natural language, the questions they'd otherwise jump between three dashboards to answer.

Ledger MCP
Read-only access to accounts, transactions, anomalies, reconciliation, maker-checker and more — the same data, exposed as agent tools.
Read-only
Grafana MCP — Observability
Query Prometheus and Loki, search dashboards, and summarise firing alerts. "Is the platform healthy right now?" answered in chat.
Observability
Wazuh MCP — SIEM
Search security alerts, summarise by severity and MITRE technique, and check agent health. "Are we under attack?" — without leaving the console.
SIEM
Gated & Audited
The observability and SIEM tools require a dedicated permission, are off by default, and every call is audit-logged and traced. Read-only — no active response is ever exposed.
Permissioned

Flo Knows the Page
You're Standing On

Every operator page carries its own embedded context and a tailored set of prompts, so the assistant is always relevant to what's in front of you — no copy-pasting screens into a chatbot.

Embedded Page Context
Each page ships metadata describing what it is and what its data means, so Flo understands the screen it was opened from before you type a word.
Tailored Starter Prompts
Per-page suggested prompts reference that page's own data, turning a blank chat box into a guided set of useful next questions.
Inline Help & "Ask Flo"
The same dual-persona documentation that powers Flo also drives inline help on each page, and "Ask Flo" links hand a specific row straight to the assistant.

Paperwork In,
Structured Data Out

KYC means documents — IDs, proofs of address, registration papers. LedgerFlow pairs Docling document extraction with an LLM to read them, so onboarding isn't a manual retyping exercise.

Docling OCR Extraction
Uploaded KYC documents are processed through Docling to extract text and structure from scans and PDFs before any decision is made.
LLM Field Capture
An LLM turns the extracted text into the structured fields the KYC workflow needs, flagging mismatches for a human reviewer rather than auto-approving.
Serialised & Controlled
Document processing is queued and run one job at a time on a schedule — predictable load, no concurrent calls hammering the extraction service.
Private by Deployment
Regulated tenants can run Docling self-hosted so documents never leave the perimeter — document intelligence without an off-shore round trip.

AI Shows Up
Everywhere It Helps

The AI layer isn't a fifth pillar — it amplifies the four you already have.

See what Flo can do with your ledger.
Start a 30-day free trial and meet your in-platform analyst. No credit card.
Start Free Trial →Book a Demo