README.md

---

license: cc-by-4.0
language:
  - en
pretty_name: Macrostack Open Alternatives
size_categories:
  - n<1K
task_categories:
  - tabular-classification
tags:
  - open-source
  - self-hosting
  - software-alternatives
  - cloud
  - licensing
  - vendor-lock-in
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/alternatives.csv
---


# Macrostack Open Alternatives

**387 verified open-source and self-hostable alternatives** to
85 proprietary products, across 78 categories and the
five layers of the AI and cloud stack.

Maintained at **[macrostack.net](https://www.macrostack.net)**. Data version `2026-09-05`.

## What makes this different from a listicle

Every licence is read from the project's own `LICENSE` file rather than from a
marketing page. That sounds like a small distinction and it is not: vendors
routinely describe source-available and business-source licences as "open
source", and a dataset that repeats the marketing is worse than no dataset.

Each row also carries what a person actually needs in order to decide:

- **`self_hostable`** — can you run it yourself at all

- **`local_first`** — does it work with no network
- **`pricing`** — the shape of the cost, not a number that rots
- **`sovereignty_score`** — a transparent 0–100 score for how much control you keep

- **`stack_layer`** — where it sits, L1 chips through L5 applications

Of 387 entries, **295 are open source** and
**311 are self-hostable**.

## Columns

| Column | Meaning |
|---|---|
| `alternative` | The open or self-hostable option |
| `replaces` | The proprietary product it is an alternative to |
| `vendor_replaced` | Who sells that product |
| `category`, `category_slug` | The decision category |
| `stack_layer`, `stack_layer_name` | L1 chips → L5 applications |
| `license` | SPDX identifier, read from the project's own LICENSE |
| `open_source` | True only for an OSI-approved licence |
| `self_hostable` | Can be run on your own infrastructure |
| `local_first` | Works without a network connection |
| `sovereignty_score` | 0–100, how much control you retain |
| `recommended` | Our pick in that category |
| `pricing` | free · freemium · paid · usage |
| `kind` | software · service · hardware |
| `website`, `repo`, `docs` | Canonical links |
| `last_verified` | When a human last checked the row |
| `source_page` | The page on macrostack.net this row is rendered on |

## Use it

```python

from datasets import load_dataset



ds = load_dataset("Macrostack/macrostack-open-alternatives", split="train")



# Self-hostable options that are genuinely open source

open_selfhost = ds.filter(lambda r: r["open_source"] and r["self_hostable"])

print(len(open_selfhost))

```

Straight from the source, always current:

```bash

curl -sL https://www.macrostack.net/data/macrostack-alternatives.csv

```

## Licence and attribution

**CC-BY-4.0.** You may use this commercially. Attribution is a condition of the
licence, and it must include a link:

> Data from [Macrostack](https://www.macrostack.net) — CC-BY-4.0

## Method and limits

The full method is at [https://www.macrostack.net/methodology](https://www.macrostack.net/methodology).

Worth stating plainly:

- **`sovereignty_score` is our judgement**, not a measurement. The inputs are

  documented; the weighting is a considered opinion and you may disagree with it.

- **`recommended` is a single pick per category** and never reflects a

  commercial relationship. Several top-ranked entries here have no affiliate

  programme at all, which is the point.

- **`pricing` is a shape, not a price.** Real prices change faster than any

  dataset can track, so we record the model instead.

- **Coverage is deliberately narrow.** Infrastructure, developer tooling, AI and

  business software. Not consumer apps, not games.



Corrections are welcome — every row links to the page it came from.