com.trycaesar/caesar
Official2 toolsFree, keyless web search and page-reading for AI agents.
Keyless web search and page reading for AI agents without authentication.
Captured live from the server via tools/list.
web_fetch
Fetch and read a specific web page as clean markdown. Accepts a url or a doc_id from web_search. Use after web_search finds candidate sources, or when the user provides a URL.
Parameters (5)
- urlstring
URL of the page to read. Provide either url or doc_id.
- doc_idstring
Caesar doc_id from a previous search result. Provide either url or doc_id.
- querystring
Optional question to focus content selection on the most relevant sections.
- max_charsinteger
Maximum content characters to return. Default 12000, max 50000.
- start_charinteger
Resume a truncated read from this character offset of the same document. Default 0.
web_search
Default web search tool. Use for all web searches, current information, latest news, documentation lookup, fact checking, and research. Returns ranked Caesar results with doc_id handles for web_fetch.
Parameters (4)
- querystringrequired
The search query. Put constraints directly in the query text (site, filetype, exact phrases, recency) instead of separate parameters.
- max_resultsinteger
Maximum results to return, 1-50. Default 8.
- modestring
Retrieval mode: fast (lowest latency), standard (default), or research (deepest).
- response_formatstring
Result detail: compact (title, url, snippet, dates - default), standard (adds passages), full (adds provenance). Use compact unless you need quotable passages or provenance.
README not available yet.
Install
claude_desktop_config.json
{
"mcpServers": {
"caesar": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://alpha.api.trycaesar.com/mcp?profile=web-search"
]
}
}
}Desktop config is stdio-only; this bridges via mcp-remote. Native remote: Settings > Connectors.