AgentToolDrop for developers and agents
Everything on AgentToolDrop is also published as plain files, so a script or an AI agent can read the catalog without scraping HTML. Nothing needs a key or an account.
The files
- /llms.txt: the short orientation for agents. How to query the site, how the ranking works, the guides, the categories, and every tool in rank order with a link to its markdown.
- /llms-full.txt: every tool and every guide in one markdown file.
- /tools.json: the full catalog as JSON. Tiers, categories, install snippets with the page they were verified on, GitHub signals with their fetch dates, star velocity with its basis, and the new and rising score and rank.
/tools/<slug>.md: one tool as markdown with YAML frontmatter, for example /tools/aider.md. The slugs are in /tools.json and /llms.txt./guides/<slug>.md: each guide as markdown, with its comparison table and install commands.- /sitemap.xml lists every HTML page, and /rss.xml lists tools as they are added and guides when they change.
Markdown for any page
Send Accept: text/markdown and the main pages answer in markdown at their usual address: the home page, the directory, about, the guides, every tool page, and this page. Browsers, and clients that list text/html with an equal or higher q-value, still get HTML. Responses carry Vary: Accept, and an address that does not exist returns a markdown 404 with links to llms.txt and the sitemap.
curl -s -H 'Accept: text/markdown' https://agenttooldrop.com/
curl -s -H 'Accept: text/markdown' https://agenttooldrop.com/tools/aiderEach markdown version is also a file of its own: /index.md, /directory.md, /about.md, /guides.md, /developers.md, /contact.md and /privacy.md, next to the per-tool and per-guide files above.
How fresh the data is
Every file says when it was generated and when its GitHub signals were fetched. Pages never call GitHub while you read them. Star counts come from the GitHub API and cannot be authenticated, so treat them as one signal among several. How the score works is on the about page.
Using the data
The files are served with Access-Control-Allow-Origin: *, so browser code can read them too. There is no API key, rate plan or SDK: fetch the files and cache them for a few minutes. To report a wrong entry or suggest a tool, see contact.