# model-ledger

> git for models — the open, agent-native source of truth that discovers every model, rule, and pipeline across all your platforms as one immutable graph.

## Start here

- [model-ledger](https://block.github.io/model-ledger/index.md): git for models — discover every model, rule, and pipeline across all your platforms as one immutable, queryable graph.
- [Quickstart](https://block.github.io/model-ledger/quickstart.md): Install model-ledger and reach your first dependency trace in under a minute — no infrastructure, no credentials.

## Concepts

- [Concepts](https://block.github.io/model-ledger/concepts/index.md): The whole model in three ideas — the DataNode graph, the Snapshot event log, and Composites.
- [Composites](https://block.github.io/model-ledger/concepts/composite.md): Governed groups whose members are themselves models — a business-level entity that rolls up its technical components, each governed in its own right.
- [DataNode & the graph](https://block.github.io/model-ledger/concepts/datanode.md): Everything is a DataNode with typed ports. Declare inputs and outputs; the dependency graph builds itself.
- [Snapshots & the event log](https://block.github.io/model-ledger/concepts/snapshot.md): A model is an identity; everything that happens to it is an immutable, content-addressed Snapshot. The inventory is an append-only log.

## Guides

- [Agents (MCP)](https://block.github.io/model-ledger/guides/agents.md): The MCP server is the product. Eight tools and three resources let any agent explore, register, and reason about your model inventory.
- [Choosing a backend](https://block.github.io/model-ledger/guides/backends.md): Storage is storage-agnostic and one line to change — in-memory, SQLite, JSON files, Snowflake, or a remote HTTP API.
- [CLI](https://block.github.io/model-ledger/guides/cli.md): The model-ledger command line — launch the MCP server or REST API over any backend, and work with a local inventory.
- [Connectors & discovery](https://block.github.io/model-ledger/guides/connectors.md): Point factory connectors at your SQL databases, REST APIs, and GitHub repos to discover models automatically — or implement the SourceConnector protocol for anything else.

## Recipes

- [Recipes](https://block.github.io/model-ledger/recipes/index.md): Copy-paste solutions to real tasks — impact analysis, point-in-time inventory, scheduled discovery, and more.
- [Recipe: Discover from a SQL registry](https://block.github.io/model-ledger/recipes/discover-sql.md): Point a connector at a database table and pull models into the ledger on a schedule, idempotently.
- [Recipe: Impact analysis](https://block.github.io/model-ledger/recipes/impact-analysis.md): Before you deprecate or change a model, walk the dependency graph downstream to find everything it would break.
- [Recipe: Point-in-time inventory](https://block.github.io/model-ledger/recipes/point-in-time.md): Reconstruct exactly which models were active, and in what state, on any past date — straight from the immutable event log.

## Reference

- [API Reference](https://block.github.io/model-ledger/reference/index.md): The public API, generated directly from the source docstrings — it can never drift from the installed version.

## More

- [Glossary](https://block.github.io/model-ledger/glossary.md): The vocabulary of model-ledger — DataNode, Snapshot, ModelRef, Composite, and the rest, in one place.
- [Governance](https://block.github.io/model-ledger/governance.md): A complete, auditable, point-in-time model inventory — the durable building blocks every model-risk regime asks for, mapped to model-ledger primitives.
- [includes/abbreviations.md](https://block.github.io/model-ledger/includes/abbreviations.md): DataNode: The core graph primitive — anything with typed input/output ports (model, rule, ETL, queue). DataPort: A named connection point on a DataNode; depe…
- [Installation](https://block.github.io/model-ledger/installation.md): Install model-ledger and just the extras you need — SDK core is tiny; surfaces and backends are opt-in.