The minimum to wire the WordPress plugin into WordVibe and start chatting with an AI that calls WordPress tools for you.
The plugin is open-source (GPL) and holds your local capability policy — which files the AI can read/write, which DB tables it can touch, etc. The server is closed and holds the orchestrator, system prompts, and LLM credentials. Every tool the agent calls runs on your WordPress host; every reasoning step runs on our server.
Relay mode (default): the plugin POSTs the goal to WordVibe, then polls. Works everywhere; the long connection is held by the WordPress host's PHP-FPM.
Browser-direct mode (cap-immune): the chat opens an EventSource straight to WordVibe's servers. The WordPress host is removed from the long-lived connection entirely, so a 120s LiteSpeed/Hostinger proxy cap can no longer kill a long agentic run. Toggle in WordVibe → Browser-direct streaming.
Your subscription tier caps which models the router can pick:
You can request a specific model in the chat composer; the server downgrades to your tier ceiling if it's outside your plan.
// health
GET /v1/health
// start an async agentic run (browser-direct mode)
POST /v1/runs/start
{ goal, context, options, tools_manifest, browser_tools:true }
→ { run_id, session_id }
// live SSE event stream for that run
GET /v1/runs/:id/stream
// deliver a tool result back (browser-direct mode)
POST /v1/runs/:id/tool_result
{ call_id, ok, result, error }
// sync run (non-browser callers; bounded by host timeout)
POST /v1/runs
{ goal, context, options, tools_manifest, callback_url, callback_secret }
Tier comparison + checkout — coming soon. While we finalise pricing, sign up free and bring your own API key.
Open the chat in WP admin and ask. The AI knows about itself. For human help: hello@wordvibe.io.