OpenHive
Compose an AI team that fits how you work.
It never sleeps. Scale it as wide as you want.
Setup
First launch walks you through connecting a model provider (OAuth or API key, required), MCP tools, and your first agent team. No cloud, no signup — runs on your machine.
npm i -g openhiveaiEverything you need, in one local app.
From the data layer to the team layer to the world outside — OpenHive ships the whole stack as a single binary.
Sketch a hierarchy of agents on a canvas. Pick a model per node. The Lead delegates by tool call — your org chart is the program.
Drag blocks onto a grid. Bind them to your local SQLite or to an external data source. Agents read and write the same store.
Give agents tools through MCP — Slack, Notion, HubSpot, Supabase, your internal services. Bring any server you run.
01 — Agent Company
Design a company of agents.
Drop agents on a canvas, wire reporting lines, and pick a model per node. The Lead delegates by calling a tool at runtime, choosing assignees from the chart you drew — no DSL, no YAML, no compiled graph. Or skip the blank canvas and pull a ready-made company, team, agent, or panel from the Frame Market.
02 — Dashboards
Build a dashboard over your data.
Pull the panels you need from the Frame Market and arrange them in any shape and position you want. Fill them with the data you want — ask AI to build a company database inside the app, or connect an external API to stream data in real time.
03 — Integrations
Plug it into the software you already use.
OpenHive speaks MCP, the open protocol for tool servers. Pick a preset like Slack, Notion, or Supabase, or attach any MCP server you run yourself — each agent only gets the tools it actually needs.
- mcpSlackmessaging
- mcpNotiondocs
- mcpHubSpotcrm
- mcpSupabasedatabase
- mcpBrave Searchsearch
- mcpTavilysearch
- customBring your ownany mcp server
npm i -g openhiveai