
by abe238
hotin ranks what is trending in AI today across repos, models, papers and news, and ships an MCP server so agents can read the same board. It also runs as a command-line tool.
PluginContact for pricing
Step 1: Run claude mcp add hotin -- uvx hotin mcp, or open the hotin.mcpb bundle in Claude Desktop.
Step 2: For other clients, add a server entry with command uvx and args hotin, mcp.
Step 3: Ask your agent which AI repos or models are worth a look this week.
Step 4: Optionally set GITHUB_TOKEN to enable the insiders signal.
Step 5: In a terminal, run hotin, hotin models or hotin brief to read the board yourself.
Two: hotin_board, which returns a chosen tab with its evidence, and hotin_brief, which returns the daily digest.
No for the core sources. Optional keys unlock Reddit and YouTube, and a GitHub token enables the insiders signal.
No. The README says it runs on your machine and hotin.ai is never in the path.
A first call on a cold cache takes about 15 seconds while it fetches. After that answers come from the local cache in about a second.
Requirements
Motif Research MCPMotif MCP turns a folder of interview transcripts into cited findings and checks them with a critic loop. It runs inside MCP hosts such as Claude Code, Cursor and Claude Desktop.
CodeCartographer MCPCodeCartographer turns a repository into layered architecture notes, behavioral contracts, defect findings and a language-agnostic reimplementation spec. It ships as a Pi extension, an MCP server and a drop-in template.