CBCharlie Barmore
← All projects
LedgerClaw logo

LedgerClaw

The AI-native, agentic operating system for the modern CPA firm: agents that keep the books, run the close, and carry the work from book to tax, orchestrated by the practitioner.

Live - waitlist For solo & small CPA firms
Join the waitlist →

All product screens on this page show demo data, never real client or user information. They’re here to demonstrate what the product can do.

LedgerClaw Agents: a catalog of installable AI workers for a CPA firm. Bank Reconciliation and Categorization Drift Detection are already active, with Bookkeeping (A/R & A/P clearing, missing-JE auto-heal), Period Close, and Quality Control agents ready to install

About

LedgerClaw is the AI-native operating system for a CPA firm: one place to onboard a client, keep their books, carry the work from book to tax, and run the practice, with a team of agents doing the heavy lifting and the practitioner in control.

  • Onboard a client in days, not weeks. The Books Review agent scrubs the prior bookkeeper’s work and flags everything that looks wrong, right when a firm feels the most pain bringing a client over. (The category pitches this as “monthly review”; LedgerClaw turns the same agent into a new-client accelerator. It’s the wedge.)
  • Run the books month to month. Plaid sync, AI categorization with a learning loop, journal entries, and period close.
  • Hold every client on one firm spine. Each engagement, time entry, tax workpaper, and Vault document lives inside a client context, with no switching tools between bookkeeping and practice management.
  • Talk to your firm. A persistent AI assistant answers reporting questions, drafts entries (with a JE preview before anything posts), logs time, and adds leads, vendors, and clients.
  • See what matters, proactively. A daily briefing and an anomaly engine surface cash-runway warnings, duplicates, stale uncategorized items, A/R aging, and more. The system tells you what to look at, not the other way around.

It runs on one multi-tenant firm spine that holds every client org in a single workspace: bookkeeping and close, practice management (engagements, pipeline, time, billing), a document vault, CRM, SMS capture, light tax workpapers, and the agents, so there’s never a switch between the books and the practice.

It’s built around a simple belief: the people who actually do the work should own the software that does the work, tools shaped by the way real CPAs run real engagements. In production today at ledgerclaw.app, behind a waitlist.

How I built it

LedgerClaw is AI-native from day one, not bolted on. Built 100% by me solo, with coding agents as the engineering team.

  • Two Claude models, two jobs. Claude Sonnet powers categorization and Vision OCR for receipts; Claude Opus runs the conversational tool-use loop.
  • Agentic workflow: investigate → propose → execute. The assistant chains tool calls, asks for confirmation, then acts. Every write tool (record an invoice or bill, create a contact or lead, log time) is preview-first; nothing hits the ledger without a confirm.
  • Nine production agents under one framework (categorization, Books Review, anomaly scan, daily briefing, outreach drafting, lead research, and more), scheduled via Railway Cron and surfaced through a shared Activity Feed.
  • Multi-tenant by construction. Every table is org-scoped, so a firm holds many client orgs and switches between them like workspaces.

Stack

FastAPIPostgres 16React 19ClerkPlaidClaude (Opus + Sonnet)Railway
The firm dashboard: revenue, expenses, net income, and a health score at a glance, a 12-month revenue-vs-expenses chart, connected accounts with their to-categorize queue, and a spending-by-category breakdown, with the full practice nav down the side.
The firm dashboard: revenue, expenses, net income, and a health score at a glance, a 12-month revenue-vs-expenses chart, connected accounts with their to-categorize queue, and a spending-by-category breakdown, with the full practice nav down the side.
AI bookkeeping: bank and card activity auto-categorized to the chart of accounts (each tagged 'Auto') and reconciled against the bank, with a short review queue for anything the agents aren't sure about.
AI bookkeeping: bank and card activity auto-categorized to the chart of accounts (each tagged 'Auto') and reconciled against the bank, with a short review queue for anything the agents aren't sure about.
Ask LedgerClaw, the built-in assistant that answers from the real books; here it pulls 2026 year-to-date revenue straight off the income statement and offers to break it down by client or month.
Ask LedgerClaw, the built-in assistant that answers from the real books; here it pulls 2026 year-to-date revenue straight off the income statement and offers to break it down by client or month.
Book to tax on one workpaper: the Working Trial Balance runs Beginning → Activity → Unadjusted → Book AJEs → Adjusted → Tax AJEs across the columns, with live balanced checks on activity, AJEs, and TJEs, accounts grouped by type, and a one-click Excel export.
Book to tax on one workpaper: the Working Trial Balance runs Beginning → Activity → Unadjusted → Book AJEs → Adjusted → Tax AJEs across the columns, with live balanced checks on activity, AJEs, and TJEs, accounts grouped by type, and a one-click Excel export.