MCP#

waper exposes MCP (Model Context Protocol) over Streamable HTTP. Your agent client can read briefs, search the library, write documents, create watches and transcribe media.

Endpoint#

Client setup#

Claude Code#

{
	"mcpServers": {
		"waper": {
			"type": "http",
			"url": "https://waper.ai/mcp",
			"headers": { "Authorization": "Bearer wk_xxx" }
		}
	}
}

Cursor / Claude Desktop#

{
	"mcpServers": {
		"waper": {
			"url": "https://waper.ai/mcp",
			"headers": { "Authorization": "Bearer wk_xxx" }
		}
	}
}

Other MCP clients#

Use Streamable HTTP with JSON responses; initialize returns the waper-mcp server info and tool list.

Tools#

Briefs and knowledge:

• search_watches, list_followed_watches: find watches;
• get_brief: one issue;
• ask_watch: ask about a watch — answers only from brief items;
• get_knowledge: a watch's knowledge base (weekly reports, entities, timeline);
• search_docs, read_doc, list_docs, create_doc, edit_doc: documents;
• search_library, save_link, read_library_item: the library;
• list_spaces: spaces;
• transcribe_media, get_transcript: media;
• create_watch: start a watch (counts toward the monthly limit);
• remember: write a fact into agent memory;
• search_media: search transcripts.

Cost & limits#

• Each call charges actual model cost in credits (search_*/read_* are a few credits; create_watch and transcribe_media show a quote first);
• Rate limits in API keys;
• Tools only touch your own data; each key is audited separately.