How i should use this?

[itspers]

Found this repo while doing research on trying to understand why and how to use new claude subagents and decided to try.

So i run 'npx claude-flow@alpha init --force' and it added tone of files to repo and override claude.md. Then i run npx claude-flow@alpha swarm, it burned some tokens for some time and finally says:

You can now assign tasks to the swarm using commands like:

• npx claude-flow@alpha task "Your task description"
• Or use the MCP tools directly for more control

So i dont supposed to use this claude instance to do something? Lets say i work today on some feature, and as part of it i need to add another 10 tables to DB. Usually i just open claude code in vscode and write 'add 10 tables in prisma schema' and it do here.

Is this a 'task' that i need to run from separate terminal? (npx claude-flow@alpha task "Your task description")?

[jbbjbb]

Just getting started myself, Rueven has a less than efficient demo of prior version here:
https://www.linkedin.com/video/live/urn:li:ugcPost:7338958277646393345/?originTrackingId=98BFbYghSVqcncNLBFxvDA%3D%3D

But after init I have been starting with this example, replacing the project prompt with a more traditional (and specific) project prompt, updating the namespace, and reducing agent number:
npx claude-flow@alpha hive-mind spawn "Build e-commerce platform with React, Node.js, and PostgreSQL"
--agents 10
--strategy parallel
--memory-namespace ecommerce
--claude

Interesting. Have reached a good endpoint with the first bit, need additional check and input loops (I wouldn't let me employees work with such wild abandon). Still prone to go off the rails - e.g. built an API end point example creator/educator for fin service, and it went off the rails and created a whole demonstration suite that doesn't make any live calls or show example JSON. Prompt window after build still allows for interactive follow up, although this is substantially more awkward in Claude Code terminal session where LFs and copy paste are more awkward. Suspect starting with a plan phase first and then launching a hive-mind or swarm may go better.

Found this repo while doing research on trying to understand why and how to use new claude subagents and decided to try.

So i run 'npx claude-flow@alpha init --force' and it added tone of files to repo and override claude.md. Then i run npx claude-flow@alpha swarm, it burned some tokens for some time and finally says:

You can now assign tasks to the swarm using commands like:

• npx claude-flow@alpha task "Your task description"
• Or use the MCP tools directly for more control

So i dont supposed to use this claude instance to do something? Lets say i work today on some feature, and as part of it i need to add another 10 tables to DB. Usually i just open claude code in vscode and write 'add 10 tables in prisma schema' and it do here.

Is this a 'task' that i need to run from separate terminal? (npx claude-flow@alpha task "Your task description")?