About · PG-23
We build the accounting data layer beneath AI-native finance
Ledgerbook AI is a headless, MCP-native, self-hostable accounting data layer — the governed book of record that finance teams, planning agents, auditors and application builders all write to and read from.
What we build
Our mission is to engineer a ground-up, AI-native, self-hostable, air-gap-ready enterprise accounting data layer that acts as the immutable, high-performance financial nervous system for modern organisations.
We are deliberately not fighting for the “AI-native accounting application” label — that space belongs to the application cohort, and applications are our buyers, not our rivals. We sell the layer underneath: one immutable, bitemporal book of record, correctable only by provable, attributable entries, deployable wherever the data has to live.
Four principles
Headless by creed
The core ships no UI and takes no workflow position: four governed surfaces — REST, GraphQL (read; write parity on the roadmap), CLI and MCP — plus streaming ingress and webhook egress are the product. Your application renders it; the ledger proves it.
NG-1 · PH-1
Append-only by construction
Final postings cannot be altered or deleted. Corrections are typed reversing, adjusting or restating entries linked to the original, and the chain hash is on by default.
FR-205, FR-211, FR-118/119 · PH-1
Provenance everywhere
Every mutation is a provenanced event with who/when/where, exportable as OCEL 2.1 and verifiable by a third party without trusting us.
FR-208/234–238, FR-905 · PH-1
Deploy anywhere
Cloud, self-hosted or air-gapped is a deployment choice, not a product tier. The offline bundle runs with zero egress — including the MCP surface.
NFR-6/8 · PH-1; embedded class, multi-region Roadmap (PH-2)
Research and open commitments
We work from dossiers: public evidence about the ledgers, agent stacks and competitors in this lane, captured on dates and cited. Where we compare ourselves to another product, the comparison says where they win and when it was last verified — that discipline is the brand, not a marketing device.
Two commitments follow from it. First, open-core: the core is source-available and the licensing text publishes before commercial launch — the stance lives on pricing & licensing. Second, RFC-2119 precision: our docs and specifications say MUST, SHOULD and MAY because those words have meanings, and marketing copy that restates them keeps the same honesty.
We publish what ships and we label what has not shipped yet. Where a capability is on the roadmap, the page says Roadmap next to it — every time.
Effective 2026-09-19 · entity: Ledgerbook AI
Work with us
Design partners, planning vendors and integrators: the programme is open.