docs: reframe README around programmable/headless control (less Claude-specific)

[gitcoder89431]

Summary

The README currently leads with "MCP (Claude integration)" and explicitly mentions Claude and ~/.claude.json, which makes the project feel tightly coupled to one AI product.

However, the actual design of the CLI + MCP server is more generally useful:

• A programmable, scriptable, headless control surface over a single Syncthing folder
• Tools that a human dev (or any automation/agent) can use via stdio, not just Claude

Proposal

Rework the README to emphasize:

1. Primary positioning

   • meshd as a programmable, scriptable, headless control surface for a self-hosted "Dropbox" folder (~/meshd)
   • TUI + CLI + stdio API for automation
   • Syncthing as the underlying transport, but meshd as the opinionated UX

2. Use-case driven sections

   • "For developers":
     • Shared ~/meshd scratchpad across machines
     • CLI flows: meshd status, meshd devices, meshd errors, meshd ignore add
     • Headless usage via meshd status --json for bar widgets / scripts
   • "For automation":
     • Document meshd mcp as a generic stdio API, not Claude-specific
     • Show high-level examples: wait for sync, add ignores, check activity
     • Frame it as "agents/automation" instead of "Claude"

3. MCP / agent section

   • Rename section from "MCP (Claude integration)" to something like "Programmable API / MCP" or "Headless control and agents"
   • Keep the Claude example, but:
     • Move it after the generic explanation
     • Phrase it as "Example: wiring into Claude via MCP" rather than the only path

4. Tone / phrasing

   • Make the README feel public-facing and dev-oriented:
     • Focus on problems solved (single-folder sync, cross-machine workflow, scriptable status)
     • Explicitly call out that tools are designed for real developers and automation, not just one AI assistant

Acceptance Criteria

• README has a "Why this exists" or "Use cases" section describing:
  • Terminal-first dev workflows with ~/meshd
  • Headless / scriptable scenarios (status, errors, ignores)
  • Agent / automation use without naming Claude as the primary target
• MCP section title and content use generic language ("agents", "automation", "MCP") and only mention Claude as an example
• No behavior changes; this is documentation-only

[gitcoder89431]

README rewritten: developer-first framing, full CLI reference including all new subcommands, MCP section reframed as headless/agent control rather than Claude-specific.