I cannot get Ruflo working

[pimottenhoff]

For some reason, I follow EVERY install line and even use a lot of my tokens to let claude explain to me how to install ruflo and get the hive-mind working. However, it does not work and I cannot seem to figure it out. How is this possible and is there ANYONE with clear instruction on how to use and set up a hive mind using VS code on Windows with Claude Code globally installed?

Any help would be greatly appreciated!

[roman-rr]

This is a common frustration. The issue isn't your installation — it's that most of what you're trying to get working doesn't have an execution backend.

We ran an independent audit of all 300+ MCP tools in ruflo v3.5.51. Here's what actually works vs what won't work no matter what you do:

Works:

• memory_store/search/retrieve — real HNSW vector search
• embeddings_generate/compare — real embedding generation
• session_save/restore — real state persistence
• terminal_execute — real shell execution

Won't work (stubs with no backend):

• agent_spawn — creates a JSON record {status: 'idle'}, never starts a process
• task_assign — fires an EventEmitter event nobody listens to
• hive-mind — the spawn command does start a real Claude process, but the 'swarm coordination' is just reading/writing a shared JSON file
• neural_train/predict — returns random accuracy, predictions always 'coder'
• wasm_agent_prompt — echoes your input back

What to do instead:

1. If you want multi-agent orchestration → use Claude Code's native Agent tool (spawns real subprocesses)
2. If you want cross-session memory → ruflo's memory_store/search actually works for this
3. If you want the hive-mind swarm → know that it's just child_process.spawn('claude', ['--dangerously-skip-permissions', '<prompt>']) under the hood

It's not your fault — the tools register with proper schemas and descriptions but have no execution logic behind them.

Full audit: Independent Audit