Where Botverse runs
Botverse is built for agents running in normal compute — connect directly and there are no special limits. One small set of constraints applies only inside hosted AI-chat assistants that sandbox network access. Know which side you're on.
Two ways an agent reaches Botverse
Direct integration — your own agent (LangChain, CrewAI, a custom build), a coding agent running locally, a server or CI job, the botverse CLI, or the REST API. It runs in normal compute with normal network access. Everything works, no caveats.
A hosted AI-chat assistant — a person chatting in a hosted assistant (Claude, ChatGPT, Gemini, and similar) that reaches Botverse through an MCP connector. These platforms sandbox outbound network access for safety, which adds a few constraints that exist between the chat and Botverse — not in Botverse itself.
If you're integrating Botverse (the common case)
None of the sandbox limits apply. Use the upload flow, public URLs, or inline content freely; connect over MCP or REST. This is how Botverse is meant to run in production — your agent calls a tool, Botverse does the compute, you get a result back.
If you're working inside a hosted AI chat
Two reliable routes when the sandbox blocks a direct upload:
1. Point at a public URL. Host the file and use convert_from_url / transcode_from_url / transcribe_from_url. Direct download links and share links from Dropbox, Google Drive, OneDrive (personal or business), SharePoint, and Box all resolve automatically to the file — pass the share URL as-is.
2. Use the botverse CLI from your own machine, where network access is normal — it uploads and runs the job directly. See the CLI guide.
For short clips (a few hundred KB), inline convert_content / transcode_content work without any upload.