PDF Parser
The PDF parser tool extracts text and tables from a PDF at a URL, returning content agents can analyse, summarise, or extract data from.
Tool: parse_pdf
Downloads a PDF from a URL and extracts its content as plain text and Markdown tables.
Arguments
| Argument | Type | Description |
|---|---|---|
file_url | string | URL of the PDF to parse (required) |
The whole document is always parsed; there is no page range or table switch.
Output
For each page, in order:
- a
--- Page N ---marker - the page's text
- the page's tables as Markdown, under a
[Tables Found on Page]line — the tables' text also appears in the page text above them
If the download or parsing fails, the call fails — Error parsing PDF: … with the reason — and a Tool step that made it fails with that reason; it is never the tool's output.
Use cases
Parse the Q3 earnings report at https://example.com/reports/q3-2024.pdf and extract all financial figures.Read the technical specification at [URL] and summarise the API changes in section 5.Extract all tables from the data sheet at [URL] and identify the column headers.Only URLs
parse_pdf fetches the file over HTTP from the SynapseOrch server, so the PDF must be at a URL that server can reach. It cannot read a local path, and it cannot read PDFs stored in the Vault — the vault keeps PDFs for download only, and its tools read text files.
A host on your own network — loopback, private or link-local addresses, including where a redirect leads — is refused unless it is listed in EGRESS_ALLOW_HOSTS. A download follows at most 5 redirects, must finish within 60 seconds, and may be at most 25 MB.
Large documents
- When a document's text is longer than the Character threshold in Settings → General (on Cloud, Large response threshold; 100,000 characters by default), the output is saved to the vault and the agent receives a reference to that file instead, which it can search with
grepor read in slices withread_file_by_lines.
Limitations
- Scanned PDFs without an OCR text layer return little or no text
- Password-protected PDFs cannot be parsed
Code Search
The code search tool gives agents semantic and lexical access to indexed repositories. It started as a single search_codebase call; it now ships as a small toolkit of eight related tools, so an agent can move from "find the file" to "open the right lines" without ever leaving the toolkit.
Excel Parser
The Excel parser tool reads .xlsx files and converts their contents to Markdown tables. Agents can then analyse, summarise, or extract data from spreadsheets without needing to execute code.