llms.txt, agent skills and Markdown docs for APIs
Answer in brief
Coding agents read documentation differently from people. An llms.txt file gives them an index of the docs in Markdown, an agent skill packages the API contract as instructions with a SKILL.md file, and Markdown versions of pages keep their context small. Check that all three describe the same contract as the OpenAPI document.
Disclosure. This site is published by Coragrid, the operator of CompanyRegistryAPI.com, and promotes it. “Best” is our editorial choice, not a measured ranking. Other products are described only from their own official pages. How we choose
llms.txt
The llms.txt proposal suggests adding a /llms.txt Markdown file to websites
to provide LLM-friendly content, at the site root or at any path within it.
For an API, it is where an agent learns what the API is, its key facts and
where the full docs are, before it reads anything long.
Read a provider’s file with the same tool your agent uses:
curl -s https://companyregistryapi.com/llms.txt | head -20
CompanyRegistryAPI.com’s file opens with a one-paragraph summary, then key
facts an agent needs to call the API correctly: no key, the rate limit and the
429 with Retry-After, the three endpoints, the search semantics, permanent
IDs, problem codes and the attribution line. It then links an agent setup
file, a lookup skill, the API docs and the OpenAPI document.
Agent skills
The Agent Skills specification describes a skill as a directory containing, at
minimum, a SKILL.md file, which must contain YAML front matter followed by
Markdown content. For an API, a skill is the contract written as instructions
a coding agent can load when the task needs it.
CompanyRegistryAPI.com publishes a lookup skill that its llms.txt describes as
the API contract in Markdown: search, company profiles, current officers,
errors and attribution. Its agent setup file tells a coding agent to check the
API, install the skill and, when the user has a Coragrid API key, connect the
cgd command line.
curl -s https://companyregistryapi.com/skills/companyregistryapi/SKILL.md | head -12
Markdown versions of pages
HTML pages carry navigation, scripts and styling an agent does not need. Markdown versions keep the text and the links. CompanyRegistryAPI.com offers two ways in:
curl -s https://companyregistryapi.com/api.md | head -20
curl -s -H "Accept: text/markdown" https://companyregistryapi.com/api | head -20
Its llms.txt adds that a company page’s Markdown version is a pointer to the record in the free API, and that the page sends every client other than a browser there.
What to check
- One contract. Do the llms.txt key facts, the skill and the OpenAPI document agree on limits, parameters and error codes? Pick three facts and compare.
- Credentials. Does the provider say which credentials apply where? CompanyRegistryAPI.com publishes an authentication note saying the API takes none and the site never asks for a key.
- Instructions an agent can act on. “Search matches the start of a name” is actionable; “powerful search” is not.
- Where agents should not go. A note like “call the API, not the HTML pages” saves an agent from scraping a page meant for people.
Your own docs
If you wrap a company data API for your team, give your wrapper the same three layers: a short llms.txt-style index, a skill with the calls and error handling, and Markdown versions of the docs. Agents then use your wrapper the way it was meant to be used.
Questions and answers
Does llms.txt replace the OpenAPI document?
No. llms.txt is an index and a summary for language models; the OpenAPI document is the precise contract. A good llms.txt links to the OpenAPI document.
How do I get the Markdown version of a CompanyRegistryAPI.com page?
Append .md to the page URL, or request the page with Accept set to text/markdown. Its llms.txt says this works for every page except search results.
Should an agent read the website or call the API?
Call the API. CompanyRegistryAPI.com's llms.txt says agents should call the API rather than read the HTML pages, whose browser search is protected and meant for people.
Sources for this entry
- The /llms.txt file
- Agent Skills specification
- CompanyRegistryAPI.com llms.txt
- CompanyRegistryAPI.com: Free Company Registry API
Each source supports only what its publisher says about its own product or data. Official descriptions do not establish comparative performance, and this site has run no benchmark. All sources
Continue reading
- MCP and WebMCP (Guide)
- Company data from the command line (Guide)
- Reading an OpenAPI document (Guide)
- The first three requests (Guide)
- Best company data API (Category guide)
- CompanyRegistryAPI.com free API (Profile)
How the related websites connect
This is one of 3 coordinated reference sites published by Coragrid. They cover neighbouring questions and feature the same product, so they are not independent endorsements of it or of each other.
Cite this entry
Best Company Data API. “llms.txt, agent skills and Markdown docs for APIs.” Content reviewed . https://bestcompanydataapi.com/llms-txt-skills-and-markdown-docs