Programmatic access
Programmatic access for your own tools.
Developer API
Mdtero API for scripts and product integrations
Use the API when Mdtero needs to plug into your own scripts, automations, or product workflows.
Best for
Internal research tools, batch pipelines, and product surfaces that need clean paper outputs on demand.
API snapshot
At a glance
Supported sources
What works best today
When needed, keep retrieval on your own machine first, then use the API for parsing and downloads.
Works best
arXiv, PMC / Europe PMC, bioRxiv / medRxiv, PLOS, Springer Open Access, and Elsevier when you have access on your own machine.
May need local browser capture
Some publisher pages, including Springer subscription pages, may ask you to finish capture locally in Chrome or Edge before parsing.
Fallback inputs
If structured full text is not available, you can still continue from a local PDF, EPUB, or another file captured on your machine.
API snapshot
At a glance
- Use an Mdtero API key for requests and keep licensed retrieval on your own machine.
- The local helper is the default path for DOI, URL, and file-based intake.
- Parse and translation tasks both return task IDs for polling and downloads.
- This page covers auth, helper install, task flow, artifact download, and optional agent handoff.
2. Base URL
Production: https://api.mdtero.com
Auth header: Authorization: ApiKey $MDTERO_API_KEY
Auth header: Authorization: ApiKey $MDTERO_API_KEY
API guide
Reference and setup flow
Keep the full API path in one place: authenticate, install the helper, submit tasks, poll status, and download the artifacts your workflow needs.
1. Authentication
All production requests need an API key from your Mdtero account.
For licensed publisher content such as Elsevier and ScienceDirect, keep retrieval on your own machine with the local helper. Use the browser extension only when a live page must be captured in a browser session.
Open Account3. Install the local helper
The normal setup is simple: download the installer, review it locally, then run it. It exposes a runnable <span class="api-inline">mdtero</span> command so DOI, URL, and file-based parsing can start from your own machine, and use the browser extension only when a supported live page requires browser capture.
4. Parse a paper
Use the local helper first for DOI or URL inputs so retrieval stays on your own machine. The helper-first production routes are <span class="api-inline">POST /tasks/parse-fulltext-v2</span> and <span class="api-inline">POST /tasks/parse-helper-bundle-v2</span>. If you already have a file, captured page, or local helper output, you can continue from that too. Keep direct <span class="api-inline">POST /tasks/parse</span> for already-clean inputs, scripts or automations, and other convenience cases rather than the default long-term path.
<strong>Response:</strong> the API returns a JSON object with <span class="api-inline">task_id</span>. The local helper may wait briefly and print the completed task payload when parsing finishes quickly.
7. Download artifacts
Parse tasks return <span class="api-inline">paper_md</span> by default, plus optional <span class="api-inline">paper_pdf</span> or <span class="api-inline">paper_xml</span>. If inline images cannot be served reliably, the task can also expose <span class="api-inline">paper_bundle</span> as a fallback ZIP. Translation tasks return <span class="api-inline">translated_md</span>.
If your coding agent should own the setup, generate a key-bound message in Account and paste it there. Keep this as an optional convenience, not the default path.