rustledger

A blazing-fast Rust implementation of Beancount

Parse and validate your ledger faster than Python beancount.

---

Why rustledger?

10-30x faster	Parse and validate large ledgers in milliseconds (see benchmarks)
No dependencies	No Python runtime, no libraries to install
Drop-in replacement	Compatible bean-* CLI commands for easy migration
Full compatibility	Parses any valid beancount file
Editor support	LSP server for VS Code, Neovim, Helix, and more
AI-ready	MCP server for AI assistants, plus ag-rledger — an agent-native CLI with JSON envelopes
Runs anywhere	WebAssembly support for browser and Node.js
Better errors	Detailed error messages with source locations
33 built-in plugins	Plus Python plugin compatibility via WASI sandbox
Bank import	CSV/OFX import with auto-detection, dedup, and categorization
Investment returns	Money-weighted (XIRR) and time-weighted returns, robust over imperfect ledgers: a built-in beangrow replacement
Budgeting	Budgeted vs actual spending from Fava-compatible custom "budget" directives, calendar-accurate pro-rating
Capital gains	Realized gains per tax lot, short vs long term, with exact @@ proceeds, calendar-correct holding periods, and optional per-lot annualized return (--irr)

Feature	rustledger	Python beancount	hledger	ledger-cli
Language	Rust	Python	Haskell	C++
Speed	Very fast	Slow	Fast	Fast
Beancount syntax	Native	Native	Via conversion	No
Query language	BQL (100% compat)	BQL	Custom	Custom
LSP server	Built-in	No	Via plugin	No
WASM support	Yes	No	Partial	No
Plugin system	Native + Python	Python	Haskell	Custom
Active development	Yes	Maintenance	Yes	Limited

When to use rustledger:

• You use Beancount syntax and want speed
• You want a single binary with no runtime dependencies
• You need LSP editor integration
• You want to use existing Python plugins

When to use Python beancount:

• You need Fava web interface (until rustledger integration)
• You have complex Python plugins with C extensions

When to use hledger:

• You prefer hledger's syntax and reports
• You need time-tracking features

Install

Platform	Command
macOS	brew install rustledger
Linux	brew install rustledger
Arch Linux	yay -S rustledger-bin
Fedora/RHEL	sudo dnf copr enable robcohen/rustledger && sudo dnf install rustledger
Windows	scoop bucket add rustledger https://github.com/rustledger/scoop-rustledger && scoop install rustledger
Cargo	cargo binstall rustledger or cargo install rustledger
Nix	nix run github:rustledger/rustledger
Docker	docker run --rm -v "$PWD:/data" ghcr.io/rustledger/rustledger /data/ledger.beancount
Binaries	GitHub Releases
npm (WASM)	npm install @rustledger/wasm
npm (MCP)	npx @rustledger/mcp-server (Model Context Protocol server)

Missing your platform? Open an issue to request it.

Coming from Python beancount? See the Migration Guide for command equivalents and plugin mapping.

AI & agents

Two complementary ways to drive rustledger from AI tools:

• MCP server — npx @rustledger/mcp-server exposes rustledger to AI assistants (Claude, Cursor, …) over the Model Context Protocol. Best for interactive assistants that call tools mid-conversation.

• ag-rledger — an agent-native CLI that mirrors every rledger command but emits a single structured JSON envelope per call instead of human-readable text, so coding agents and scripts parse results directly (no output scraping). Each envelope carries ok, a typed exit_code, the command's output under result, and a HATEOAS next_action hint; pass the inner command's --json to get fully structured result.data. Best for headless agents and shell pipelines. It ships beside rledger in the release archives (or build it with cargo build -p rustledger --bin ag-rledger --no-default-features --features ag-rledger):

  $ ag-rledger check --json ledger.beancount
  {"ok":true,"exit_code":0,"result":{"data":{"error_count":0}},"next_action":"…"}
  

  Same subcommands as rledger (check, query, format, report, extract, …) — only the I/O is JSON.

Quick Start

rledger check ledger.beancount
rledger query ledger.beancount "SELECT account, SUM(position) GROUP BY account"

This is the beginning of the README. Read the rest at the original. Read the full README ↗