Tools reference

Once connected, your agent has these LLMtoMD tools. You don't call them directly — you ask your agent in plain English and it picks the right one. This page lists what each does so you know what's possible (and what to say).

Tools fall into three groups: add documents, read & retrieve, and organize & remember.

---

Add documents

convert_url

Convert a web page or a publicly reachable file URL into clean Markdown and store it.

Input	Required	Notes
url	yes	Web page, or a PDF/Office/image file URL.
crawl	no	If true, follows same-origin links and converts the whole site (bounded).
ocr_langs	no	OCR language hint for scanned files, e.g. eng+fra.

"Convert https://example.com/docs/api to Markdown and save it."

"Crawl https://example.com/handbook and add the whole site to my library."

convert_file

Upload a small local file (≤ 8 MB) by passing its bytes inline, and convert it. Best for a short PDF, an image/scan, or a .docx.

Input	Required	Notes
filename	yes	Used for display and to infer the file type.
content_base64	yes	The file's bytes, base64-encoded (your agent handles this).
content_type	no	MIME type; inferred from the filename otherwise.
ocr_langs	no	OCR hint for scans, e.g. eng+deu.

"Convert ./report.pdf in this folder to Markdown and store it."

upload_file + convert_upload

For large or binary files (big PDFs, audio, video). upload_file returns a one-time upload URL so the bytes go straight to storage — never through the model's context — then convert_upload runs the conversion.

Tool	Input	Notes
upload_file	filename, size_bytes?, content_type?	Returns an upload_id, an upload_url, and a ready-to-run curl.
convert_upload	upload_id, ocr_langs?	Converts the uploaded file → Markdown + document_id.

"Upload ./recording.mp3 to LLMtoMD and convert it."

(An agent with shell access does the three steps for you. See Upload files.)

save_note

Save text or Markdown straight into your library — no conversion. Two uses: ingest a text document you already have (README, spec, PRD, .md), or record a decision/summary as memory that persists into later sessions.

Input	Required	Notes
title	yes	Short title (used as the filename).
content	yes	The Markdown body to store.
collection	no	File it under a project (name or id).
append_to	no	A document_id to append to instead of creating a new note.

"Save this README into my LLMtoMD library."

"Note in LLMtoMD that we chose Postgres over Mongo because of the relational joins."

---

Read & retrieve

list_documents

List your stored documents, most recent first. Pass collection to scope to one project.

"What's in my LLMtoMD library?"

get_document

Fetch a stored document's full Markdown by its document_id.

"Open the API spec from my library and summarize the auth section."

search_documents

Semantic search across your documents; returns the most relevant passages with scores. Optionally scope to a single document_id or a collection. k controls how many passages (default 5, max 20).

"Search my docs for how we handle webhook retries."

ask_documents

Answer a question using your documents (retrieval-augmented) — returns a cited answer with its sources. Scope to a document_id or collection if you want.

"According to my stored spec, what's the max upload size on the Pro plan?"

The difference: search finds passages; ask gives you a synthesized answer with citations. See Search & ask.

---

Organize & remember

list_collections

List your collections (projects) and how many documents each holds. A collection groups a project's documents (FRD, specs, design docs) so you can scope search/ask/list to it.

"List my LLMtoMD projects."

load_project (prompt)

A prompt — not a tool — that pulls a whole project's documents in as context at the start of a session.

"Load my 'Billing service' project from LLMtoMD."

See Organize with collections.

---

Notes that apply to every tool

• Auth & metering happen on the backend using your connection key — credits, rate limits, and your plan's file-size limit are enforced server-side.
• Conversions are credited. search, ask, list, and reading documents are cheap or free; converting files consumes credits per your plan.
• Long conversions return a "still processing" handle; the document appears in list_documents shortly after.