Windsurf's Cascade agent leans heavily on project files for context. Drop a converted .md into your repo and the agent reads your spec, API doc or design brief as if you'd written it yourself.
Cascade's strength is autonomy across a whole repo — it picks files, reads them, plans, edits. But that flow only works on text it can read. A PDF in your project folder is opaque; the agent skips it or skims metadata at best. Convert the PDF to Markdown and the same document becomes searchable, citeable context. For teams shipping against external specs (an OpenAPI doc as PDF, a third-party SDK guide, a vendor security policy), this is the cleanest way to make Cascade follow the spec instead of guessing.
Drop the PDF on t0md, download the .md, save it under your project (e.g. `docs/api-spec.md`). Open Windsurf, point Cascade at the file or just describe the task — the agent will discover the .md during context-gathering. For MCP-style integration, Windsurf supports MCP servers; configure t0md as a server and Cascade can call the conversion tool directly when needed.
Not as first-class context. PDFs sit in your project folder but Cascade treats them as binary attachments. Markdown is plain text, indexed, searchable and citable.
The workflow is essentially the same. Convert once, commit the .md, reference it in your prompts. Both editors index Markdown by default; both let you @-reference specific files in chat.
Yes if you've added t0md as an MCP server. Cascade can then call convert_to_markdown on any PDF path inside the conversation, no manual step.