Ollama shared via Tailscale: per-machine URL detection + Howard access

CLAUDE.md: Ollama section rewritten. localhost for Mike's workstation,
100.92.127.64:11434 via Tailscale for all other machines. Claude reads
identity.json hostname to determine which URL to use. Firewall rule
restricts to Tailscale 100.0.0.0/8 subnet only.

ONBOARDING.md: updated Ollama section for remote access.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

.claude/ONBOARDING.md

@@ -154,9 +154,16 @@ Ollama runs AI models locally on your GPU. Used for tasks that don't need Claude
 - `codestral:22b` — code generation assistance
 - `nomic-embed-text` — embeddings for semantic search
 
-**Status:** May not be installed/running on your machine yet. If Claude says "Ollama not available", it's fine — it falls back to doing the work itself.
+**Ollama runs on Mike's workstation** and is shared via Tailscale. You don't need to install it locally.
 
-**To check:** `curl http://localhost:11434/api/tags` — if it returns models, Ollama is running.
+**To use from your machine (Tailscale must be connected):**
+```bash
+curl -s http://100.92.127.64:11434/api/tags
+```
+
+If that returns models, you're connected. Claude automatically uses the right URL based on which machine you're on (reads from `identity.json`).
+
+If it fails: check that Tailscale is connected (`tailscale status`) and Mike's workstation is online.
 
 ### GrepAI (semantic code search)