Model Context Protocol · bring your own agent

Connect any AI
to your books.

LedgerBeaver speaks MCP — the open standard for connecting an AI agent to real tools and data. Plug in Claude, ChatGPT/Codex, Cursor, Gemini, a local model, or anything else that speaks the protocol. One endpoint, one key, your AI runs the numbers.

Connect MCP

Three steps. Any agent.

One endpoint, one key, one paste. The full ready-to-paste config for your client lives inside the app.

1

Copy the endpoint

This is the single URL every MCP client connects to — same endpoint for every agent, every tier.

MCP endpoint
https://ledgerbeaver-production.up.railway.app/mcp
2

Sign up & generate your key

Create a free account, then open Connect AI in the app and generate a company-scoped API key. It’s the only credential your agent needs.

Sign up free →

OAuth one-click sign-in (no copy-pasted key) is coming — keys work today with every client below.

3

Paste into your agent

Full ready-to-paste configs for every client — Claude, ChatGPT/Codex, Cursor, Gemini CLI, and more — are inside the app under Connect AI, pre-filled with your key.

Get started →
Works everywhere MCP works

One server. Every agent you'd want.

Real, verified marks for clients we've tested. Anything that speaks MCP over SSE works.

ClaudeDesktop, Code, Web
CodexOpenAI · CLI + MCP
Cursorvia .cursor/mcp.json
Gemini CLI+ Vertex AI
OC
OpenCodeopen-source agent
OW
Openclawopen-source agent
HM
Hermes Agentautonomous workflows
Your own agentany MCP-speaking client
No AI yet?

Three ways to get one.

You don't need to be technical. Pick the path that matches how hands-on you want to be.

Easiest

claude.ai

A chat app in your browser, no install. Add LedgerBeaver as a custom connector and start asking it to work your books. Paid plans start around $20/mo.

Open claude.ai →
Most capable

Claude Code / CLI agents

A terminal-based coding agent that also handles multi-step accounting workflows well — three-way matching, batch approvals, month-end. One command to connect.

Claude Code docs →
Max privacy

A local model

Run a model on your own hardware with Ollama, LM Studio, vLLM, or llama.cpp — they serve the model only. Connect it through an MCP-capable agent like Continue, OpenCode, or Aider, which bridges your local model to LedgerBeaver.

See OpenCode →

Using Claude Code, Codex, or another CLI agent? Your account’s Connect AI page has one-line setup commands for each.

Sign up to get yours →