# Model Ledger
**Category:** ๐ป Developer Tools
**Repository:** https://github.com/block/model-ledger
**Views:** 0
**Installs:** 0
**Upvotes:** 0
**Directory Page:** https://allmcps.com/mcp/model-ledger
## Description
Model inventory and governance: trace deployed models and their dependencies as an append-only log
## Claude Desktop Quick Installation
Heuristic fallback โ verify the package name and runner against the repository README before running it. Uses `npx` (confidence: low):
```json
"mcpServers": {
"model-ledger": {
"command": "npx",
"args": ["-y","model-ledger"]
}
}
```
## Documentation & README
# model-ledger
**git for models** โ know what models you have deployed, where they run, what they depend on, and what changed.
[](https://github.com/block/model-ledger/actions/workflows/ci.yml)
[](LICENSE)
[](https://python.org)
[](https://pypi.org/project/model-ledger/)
[](https://pypistats.org/packages/model-ledger)
[](https://block.github.io/model-ledger/)
๐ **[Documentation](https://block.github.io/model-ledger/)** ·
[Quickstart](https://block.github.io/model-ledger/quickstart/) ·
[Concepts](https://block.github.io/model-ledger/concepts/) ·
[Governance](https://block.github.io/model-ledger/governance/)
---
model-ledger is a model inventory for any organization with deployed models. It
**discovers** models, heuristic rules, and ETL across your platforms, maps the
**dependency graph** automatically, and records **every change as an immutable event**.
Unlike registries tied to a single platform (MLflow, SageMaker, W&B), it spans all of
them โ as one connected graph โ and it's built to be driven by AI agents through a
native MCP server.
Benchmarked at production scale: full inventory reconstruction over a ledger of 28.8k
models and 212k events runs in under a second ([CHANGELOG, v0.7.4](https://github.com/block/model-ledger/blob/HEAD/CHANGELOG.md)).
## Install
```bash
pip install model-ledger
```
## The graph builds itself
Every model is a `DataNode` with typed input and output ports. When an output port name
matches an input port name, `connect()` creates the dependency edge โ no hand-wiring.
```python
from model_ledger import Ledger, DataNode
ledger = Ledger()
ledger.add([
DataNode("segmentation", platform="etl", outputs=["customer_segments"]),
DataNode("fraud_scorer", platform="ml", inputs=["customer_segments"], outputs=["risk_scores"]),
DataNode("fraud_alerts", platform="alerting", inputs=["risk_scores"]),
])
ledger.connect()
ledger.trace("fraud_alerts")
# ['segmentation', 'fraud_scorer', 'fraud_alerts']
```
Every mutation is recorded as an immutable **Snapshot** โ an append-only event log that
gives you full history and point-in-time reconstruction, because nothing is overwritten.
## Talk to your inventory
The MCP server is a first-class surface โ point Claude (or any MCP agent) at it:
```bash
pip install "model-ledger[mcp]"
claude mcp add model-ledger -- model-ledger mcp --demo
```
> **You:** if we deprecate `customer_features`, what breaks?
>
> **Claude:** 3 models consume it directly, 2 more transitively.
## Documentation
Everything lives at **[block.github.io/model-ledger](https://block.github.io/model-ledger/)** โ
and it can't drift, because the API reference is generated from source and every example
runs in CI:
- **[Quickstart](https://block.github.io/model-ledger/quickstart/)** โ install to your first dependency trace in 60 seconds
- **[Concepts](https://block.github.io/model-ledger/concepts/)** โ DataNode, Snapshot, and Composite, in three ideas
- **[Agents (MCP)](https://block.github.io/model-ledger/guides/agents/)** โ the eight-tool agent surface, with a worked transcript
- **[Connectors](https://block.github.io/model-ledger/guides/connectors/)** โ discover from SQL, REST, GitHub, or your own platform
- **[Backends](https://block.github.io/model-ledger/guides/backends/)** โ in-memory, SQLite, JSON, Snowflake, or remote HTTP
- **[Governance](https://block.github.io/model-ledger/governance/)** โ how the primitives map to SR 11โ7/SR 26โ2, the EU AI Act, and NIST AI RMF
- **[API reference](https://block.github.io/model-ledger/reference/)** โ generated from the source
## Architecture
```mermaid
flowchart LR
subgraph Sources
C1[SQL / REST / GitHub / Prefect
connectors]
end
subgraph Core
L[Ledger
append-only event log,
point-in-time reconstruction]
G[Dependency graph]
V[Compliance profiles
SR 11-7/SR 26-2 ยท EU AI Act ยท NIST AI RMF]
end
subgraph Surfaces
S1[Python SDK]
S2[CLI]
S3[REST API]
S4[MCP server ยท 8 tools]
end
B1[(in-memory ยท SQLite ยท JSON ยท
Snowflake ยท remote HTTP)]
C1 --> L
L --> G
L --> V
L --- B1
S1 --> L
S2 --> L
S3 --> L
S4 --> L
```
## For organizations
The OSS core handles discovery, graph building, change tracking, storage, the agent
protocol, and compliance validation โ the SR 11โ7/SR 26โ2, EU AI Act Annex IV, and
NIST AI RMF profiles ship in `model_ledger.validate`. Your internal package provides
only the thin layer on top: connector configs, custom connectors for internal
platforms, and credentials. Thin config, not reimplemented logic.
## Contributing
See [CONTRIBUTING.md](https://github.com/block/model-ledger/blob/main/CONTRIBUTING.md).
All commits require DCO sign-off.
## Security
See [SECURITY.md](https://github.com/block/model-ledger/blob/HEAD/SECURITY.md) for how to report vulnerabilities privately.
## License
Apache-2.0. See [LICENSE](https://github.com/block/model-ledger/blob/main/LICENSE).
Created and maintained by [Vignesh Narayanaswamy](https://github.com/vigneshnarayanaswamy) at Block.