📚 Openapi® LLM Wiki

An LLM-optimized knowledge base to ground any AI agent on the Openapi® API marketplace

A structured, LLM-optimized knowledge base for the Openapi API marketplace.
Drop the contents of knowledge/ into any LLM context to ground it with authoritative,
up-to-date information about Openapi's company profile, platform architecture, API catalog, service
endpoints, and frequently asked questions.

How to use

As LLM context (recommended)

This repository is designed to be consumed directly by an LLM agent or tool. Choose one of the
following approaches, depending on your workflow:

1. Chat-based LLMs (Claude, ChatGPT, Gemini, etc.)
   Copy and paste the entire contents of the knowledge/ folder into a new chat or project. This
   gives the model a comprehensive understanding of every Openapi service, endpoint, authentication
   model, billing mechanism, and platform convention.

2. Agent frameworks (OpenAI Assistants, LangChain, LlamaIndex, etc.)
   Point your agent's vector store or file search tool at this repository. The markdown files in
   knowledge/ are already structured for retrieval: each service has a compact endpoint table and
   each OAS spec is available as raw JSON for tool-based API generation.

3. MCP server / custom tool
   Clone this repo and expose it through an MCP resource or a custom file-read tool so the LLM can
   selectively load the files it needs at runtime.

4. Openai-compatible providers with file upload
   Upload the entire knowledge/ tree as a project or knowledge base attachment.

5. llms.txt
   llms.txt indexes every file following the llms.txt
   convention, and llms-full.txt packs the whole knowledge base into a single
   file, ready to paste or fetch:
   https://raw.githubusercontent.com/openapi/openapi-llm-wiki/main/llms-full.txt

As an MCP server (GitMCP)

GitMCP serves this repository as a remote MCP server, so any MCP client can
search and fetch the wiki on demand (it reads llms.txt first):

https://gitmcp.io/openapi/openapi-llm-wiki

Claude Code:

claude mcp add --transport http openapi-llm-wiki https://gitmcp.io/openapi/openapi-llm-wiki

Cursor, Windsurf, VS Code and other clients (mcp.json):

{
  "mcpServers": {
    "openapi-llm-wiki": {
      "url": "https://gitmcp.io/openapi/openapi-llm-wiki"
    }
  }
}

You can also ask questions about the wiki in plain language on
DeepWiki.

Lightweight browsing

You can also browse the files manually to look up specific services, endpoints, or platform
information without an LLM. Start with knowledge/README.md which acts as
an index.

Service lookup pattern

When the LLM needs information about one specific service, load only the relevant files:

knowledge/oas/<service>.openapi.json       →  the raw OpenAPI 3 specification
knowledge/services/<service>.md            →  human-readable endpoint summary

Always pair with knowledge/platform-guide.md for cross-cutting
concerns (auth, billing, async patterns) and knowledge/faq.md for
troubleshooting.

Repository structure

openapi-llm-wiki/
├── knowledge/
│   ├── README.md                 ← index of the knowledge base
│   ├── company-profile.md        ← who Openapi is, history, certifications, data sources
│   ├── services-catalog.md       ← catalog of all 28 APIs with base URLs and categories
│   ├── platform-guide.md         ← authentication, scopes, billing, sandbox, async patterns
│   ├── faq.md                    ← console FAQ (account, API usage, payments)
│   ├── references.md             ← official websites, GitHub repos, SDKs, Postman, status page
│   ├── services/                 ← per-service endpoint references (generated from OAS specs)
│   │   ├── company.md
│   │   ├── risk.md
│   │   ├── smsv2.md
│   │   └── ... (29 service files)
│   └── oas/                      ← snapshot of official OpenAPI 3 specifications
│       ├── 00-list.txt           ← canonical URLs for all specs
│       ├── company.openapi.json
│       ├── risk.openapi.json
│       └── ... (29 OAS files)
├── huggingface/
│   └── README.md                 ← Hugging Face dataset card
├── scripts/
│   ├── build-llms.sh             ← regenerates llms.txt and llms-full.txt
│   └── build-hf-dataset.py       ← builds the Hugging Face dataset folder
├── llms.txt                      ← llms.txt index of the knowledge base
├── llms-full.txt                 ← the whole knowledge base in one file
├── CITATION.cff
├── LICENSE
└── README.md                     ← this file

What the LLM can answer

With this knowledge base loaded, the LLM can accurately answer questions such as:

• "How do I authenticate to the Openapi platform?"
• "What is the base URL for the Company API and what endpoints does it expose?"
• "Which service should I use for Italian digital signatures?"
• "How does the async request pattern work for document retrieval?"
• "What are the differences between wallet-based and subscription billing?"
• "How do I create a token via the OAuth v2 API?"
• "Is the old OAuth v1 still supported?"
• "What data sources does Openapi rely on for company information?"
• "Generate me the curl command for looking up a German company."

Keeping it up to date

The canonical source of truth for OpenAPI specifications is:

https://console.openapi.com/oas/en/<service>.openapi.json

The full list of spec URLs is in knowledge/oas/00-list.txt.

To refresh this knowledge base:

1. Re-download each spec from its canonical URL into knowledge/oas/. The
   Refresh OpenAPI specs workflow does this every Monday
   and opens a pull request when something changed.
2. Regenerate the endpoint summaries in knowledge/services/ from the updated specs.
3. Run ./scripts/build-llms.sh to regenerate llms.txt and llms-full.txt (CI fails if they
   are stale).
4. Review knowledge/services-catalog.md and
   knowledge/platform-guide.md for any platform-level changes.

The API Library in the Openapi console is the authoritative
source for which APIs are active, deprecated, or newly introduced.

Contributing

Contributions are always welcome! Whether you want to report bugs, suggest new features, improve documentation, or contribute code, your help is appreciated.

Authors

• Francesco Bianco (@francescobianco)
• Openapi Team (@openapi-it)

Partners

Meet our partners using Openapi or contributing to this project:

• Blank
• Credit Safe
• Deliveroo
• Gruppo MOL
• Jakala
• Octotelematics
• OTOQI
• PWC
• QOMODO S.R.L.
• SOUNDREEF S.P.A.

Our Commitments

We believe in open source and we act on that belief. We became Silver Members
of the Linux Foundation because we wanted to formally support the ecosystem
we build on every day. Open standards, open collaboration, and open governance
are part of how we work and how we think about software.

License

This project is licensed under the MIT License.

The MIT License is a permissive open-source license that allows you to freely use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the software, provided that the original copyright notice and this permission notice are included in all copies or substantial portions of the software.

For more details, see the full license text at the MIT License page.