LocalLLaMA weighs always-on agent stacks

The interesting bit is not the post itself, but the narrowing user expectation: local agents are being judged less as demos and more as always-on personal infrastructure.

• –Hermes Agent appears positioned as the lower-friction local option, with official Ollama setup docs, auto-detected models, memory, skills, and messaging integrations.
• –OpenClaw has stronger ecosystem visibility and Product Hunt traction, but its full-system-access model keeps attracting security scrutiny around skills, credentials, and local automation.
• –Qwen3.6-27B gives homelab users a plausible new local model target, but tool-calling reliability and context management still matter more than raw model availability.
• –The “launcher” question is a real adoption bottleneck: whoever makes Ollama setup boring, reversible, and observable wins a lot of casual agent users.