HivemindOS manual
MiroShark Companion Integration
MiroShark runs beside HivemindOS as a simulation engine. Use this page to install and connect a local copy. HivemindOS still runs your real fleet; MiroShark only rehearses.
Install and start it
Open the Swarm view. It shows one of four states:
- Connected: open MiroShark or view setup details.
- Detected: MiroShark is installed but stopped. Start it from the dashboard.
- Missing: install and start it from the dashboard.
- Needs configuration: the view names the missing prerequisite or API key.
The one-click install uses git, uv, python3.11, docker, and screen. It starts Neo4j in Docker and runs the MiroShark backend in the background. If an OpenRouter, OpenAI, or generic LLM API key is already available to HivemindOS, it is written into MiroShark’s private settings file for you.
Manual setup commands are listed in the setup details as a fallback.
Use a custom location
HivemindOS finds common install locations on its own. If MiroShark lives elsewhere, set one of these:
MIROSHARK_HOME=~/code/MiroShark
MIROSHARK_BASE_URL=http://127.0.0.1:5001
By default, MiroShark serves its web app at http://localhost:3000 and its API at http://localhost:5001, with API docs at /api/docs.
Launch from chat
Type this in dashboard chat to start a quick simulation:
/swarm-sim <scenario>
HivemindOS queues the run and posts a status link back to the chat. If MiroShark is missing or stopped, the reply explains what to fix. Use the Swarm view for templates, round and platform settings, past runs, analysis, and publishing.
What it is for
Use MiroShark to:
- test a plan before real execution;
- see where agents are likely to disagree;
- bring back summaries, stance splits, and risk signals.
Task ownership, handoffs, votes, and real execution stay in HivemindOS.