# olud.ai

> A directory of open-source AI projects, models and tools that measures the
> relationships between them: shared technical lineage, price gaps, geography,
> and monthly retrospectives. 10,000+ projects and 230 models, rebuilt every
> night from GitHub, Hugging Face and provider pricing. Free to read, no
> account, no key.

Everything below is public. No authentication is required to read any page.
Every page has a Markdown twin — send `Accept: text/markdown`, or replace
`.html` with `.md`. It is typically a third of the HTML weight.

## What makes this source different

Most directories list what exists. This one measures what connects:

- **DNA families** — which repositories share technical lineage, derived from
  rare GitHub topics they have in common. Sharing `ggml` means something;
  sharing `ai` does not. Every relationship states the topic that justifies it.
- **Monthly archives** — what was climbing three months ago, and what has since
  stopped. Built from our own daily star readings, not reconstructed.
- **Agent readiness** — whether AI tools are themselves readable by AI agents,
  measured from their own public sites.
- **Geography** — where projects come from, based on what their GitHub owners
  declare. About half do; the pages say so.

## Core pages

- [Home](https://olud.ai/): entry point, live counters and search
- [Projects](https://olud.ai/projects.html): 10,000+ tracked open-source AI projects
- [AI Tools](https://olud.ai/tools.html): curated tools with categories and comparisons
- [Models](https://olud.ai/leaderboard.html): 230 models with live pricing and benchmarks
- [Latest](https://olud.ai/latest.html): what changed today
- [Emerging](https://olud.ai/emerging.html): projects gaining traction now
- [Releases](https://olud.ai/releases.html): recent releases across tracked projects

## Measured relationships

- [DNA families](https://olud.ai/dna.html): 1,100+ technical families
- [Languages](https://olud.ai/languages.html): projects by programming language
- [Countries](https://olud.ai/countries.html): where open-source AI is built
- [Archive](https://olud.ai/archive.html): monthly retrospectives
- [Agent readiness](https://olud.ai/agent-ready.html): which AI tools speak to agents

## Comparisons

- [Model comparisons](https://olud.ai/models-vs/): side-by-side price and benchmarks
- [Tool comparisons](https://olud.ai/vs/): open-source tools compared
- [Alternatives](https://olud.ai/alternatives/): open-source alternatives to proprietary tools

## Data and API

- [API documentation](https://olud.ai/api.html): endpoints, quotas, examples
- [API catalog](https://olud.ai/.well-known/api-catalog): RFC 9727 discovery
- [Agent skills](https://olud.ai/.well-known/agent-skills/index.json): what an agent can do here
- [Authentication](https://olud.ai/auth.md): API keys only, no OAuth, no token endpoint
- [Sitemap index](https://olud.ai/sitemap-index.xml): every indexable page

## Method and limits

- [Methodology](https://olud.ai/methodology.html): what is measured, and what is not
- [About](https://olud.ai/about.html)
- [Contact](https://olud.ai/contact.html)

Stated limits, so they are not discovered the hard way:

- Star counts and health are refreshed nightly; intraday changes are not shown.
- Geography covers roughly half the catalogue — the half whose owners declare a
  location. An organisation declares its headquarters, not where a project was
  written.
- Monthly archives require at least twenty daily readings; partial months are
  labelled as such.
- We publish verifiable facts, not composite scores. There is no single
  "quality rating" here, deliberately.

## Attribution

Scores and indices are computed by olud.ai. Contextual facts come from
public sources: GitHub, Hugging Face, OpenRouter, Artificial Analysis, PyPI,
NPM, Docker Hub. Please keep the attribution when citing or redistributing.

## Optional

- [Blog](https://olud.ai/blog.html)
- [Stack builder](https://olud.ai/stack-builder.html)
- [Terms](https://olud.ai/terms/)
- [Privacy](https://olud.ai/privacy-policy/)