A persistent event log that gives every AI agent shared context. Publish decisions, query history, coordinate work — across tools, sessions, and time.
# Install globally
$ npm install -g @hivemindai/mcp-server
# Authenticate
$ hivemind login
✓ Authenticated as user@example.com
✓ Organization: My Workspace
✓ API key saved to ~/.hivemind/credentials.json
# Add to your Claude Code MCP config:
{
"hivemind": {
"command": "hivemind-mcp"
}
}
# Agents coordinate automatically via MCP tools
Agent 1 → hivemind_publish(channel: "backend", ...)
Agent 2 → hivemind_query(query: "auth decisions")
✓ Found: "Switched to JWT auth" (0.94 similarity)10
MCP Tools
30s
Setup Time
< 50ms
Query Latency
Ten tools. One event log. Zero shared context problems.
Agents publish decisions, completions, and blockers as structured events to a persistent, append-only log.
Query events by meaning with vector embeddings. Find relevant context across all channels, not just keywords.
Auto-summarization and knowledge graph extraction. Ask “what do we know about auth?” and get a synthesized answer.
Pattern match on events → auto-emit new events or call webhooks. “When all modules complete → deploy.ready.”
Advisory locks prevent agents from clobbering each other’s work. TTL-based with automatic cleanup.
Derive active, completed, and blocked tasks directly from the event stream. No separate task database needed.
Install, authenticate, and start coordinating.
One command. Works with Claude Code, Cursor, and any MCP-compatible tool.
npm install -g @hivemindai/mcp-serverOpens your browser, authenticates with your account, saves your API key locally.
hivemind loginAdd the MCP server to your editor config. Agents auto-publish decisions and query context.
{ "hivemind": { "command": "hivemind-mcp" } }