AI Markdown Agent
Your agent.
Real documents.
One prompt connects the pipeline — files in, clean Markdown out.
Delivery ready
Verified output links appear here after conversion.
Optional: download audit files for verification or troubleshooting.
Delivery ready
Markovo gives AI agents a bounded, reviewable way to turn PDFs, Office files, images, audio, video, and public pages into clean Markdown — through remote MCP with OAuth, a local stdio server, the CLI, or the REST API.
Paste one prompt into your agent and it configures the connection itself — sign-in, API key placement, and a live verification included.
The strip above copies a prompt that points your agent at markovo.net/install.md — a machine-readable playbook covering method selection, account sign-in, API key placement, and verification. No manual doc-reading required.
Clients that support remote MCP servers register https://markovo.net/mcp and complete a standard sign-in plus consent — no API key is created or pasted. Agents can convert public URLs and inspect owned jobs, usage, and billing guidance.
For local folders, the stdio server pairs markovo-mcp with MARKOVO_API_KEY and a dedicated MARKOVO_MCP_ROOT boundary, so the agent only touches the directory you choose.
The markovo CLI and the multipart POST /v1/convert API share the same estimate-first contract: every job carries a max_credits ceiling, and results arrive as reviewable Markdown bundles.
Web, API, CLI, and MCP all call the same conversion capabilities with the same Credit balance, ownership checks, and history — so an agent's output matches what a user sees in the browser.
PDF, DOCX, PPTX, XLSX, EPUB, and CSV convert with recoverable headings, tables, links, and images preserved in a page-aware bundle.
Image OCR and media transcription produce searchable Markdown with evidence files, so agents can quote results instead of guessing.
URL, GitHub, Notion, Google Docs, YouTube, and Hacker News imports turn explicitly authorized public pages into clean Markdown for agent context.
Agent-driven conversion stays inside explicit limits you set before any work runs.
Every request declares max_credits up front. The estimate endpoint reports the maximum first; jobs never exceed the approved amount, and failures or cancellations charge zero.
Results are bound to the account that created them. Agents receive scoped keys, short-lived asset links, and no access to other users' files.
Bundles include source maps, quality reports, and warnings so a person can verify what the agent produced before it feeds downstream work.
Connection, cost, and capability boundaries.
An AI Markdown agent is an assistant that can call a conversion tool to turn files and pages into Markdown. Markovo exposes that tool over remote MCP, local stdio MCP, a CLI, and a REST API — all on the same job contract.
Copy the one-line prompt on this page into your agent. It fetches markovo.net/install.md, recommends the best connection for your client — remote MCP OAuth, local MCP, CLI, or API — walks you through sign-in and API-key placement, then verifies the setup.
Remote MCP uses OAuth sign-in and needs no key. Local MCP, the CLI, and the REST API use an mk_live_ API key created under API keys in your account — the agent tells you exactly where to paste it.
No. Every conversion requires an explicit Credit ceiling before work starts. Failed or cancelled jobs charge zero, and the estimate endpoint reports the maximum before anything runs.
Any MCP-capable client can use the remote server or local stdio server; any environment that can run Python can use the CLI; any backend can call the REST API. The prompt flow detects what your client supports and recommends accordingly.