docs: add AgentOpus MCP to generative video lane

docs/AGENT-OPUS-MCP.md

@@ -0,0 +1,102 @@
+# AgentOpus MCP — generative video
+
+Hosted MCP for **Agent Opus** (OpusClip's text/script → finished video product).
+This is a **generation** surface, same family as Rafiki images + Floyo clips —
+not the OpusClip long-form → shorts flywheel in `kk-kb/tools/media_pipeline/`.
+
+Product page: https://agent.opus.pro/mcp  
+App: https://agent.opus.pro  
+FAQ: https://help.opus.pro/agent-opus/article/ao-faq  
+Prompt guide: https://help.opus.pro/agent-opus/prompt-guide
+
+## When to use which
+
+| Need | Tool |
+|------|------|
+| Still images (Gemini / OpenAI, styles, batches, archive) | **Rafiki** local MCP (`MCP.md`) |
+| Short ComfyUI / Floyo morphs, lip-sync, dry-run-first video | **Floyo** (`FLOYO.md`) |
+| Narrative explainer / promo / faceless video from prompt or script | **AgentOpus MCP** (this doc) |
+| Clip a meetup / YouTube long-form into shorts + schedule | **OpusClip** in kk-kb (`tools/opusclip_pipeline/`, `tools/media_pipeline/OPUS-MCP.md`) |
+
+Do **not** use AgentOpus to caption or cut existing footage — that is OpusClip.
+
+## Endpoint
+
+```text
+https://api.opus.pro/api/agent-mcp
+```
+
+Streamable HTTP + OAuth. No API key. Connection is scoped to the signed-in
+AgentOpus account (styles, voices, brand anchors, credits).
+
+## Install
+
+### Cursor
+
+Project or user `mcp.json`:
+
+```json
+{
+  "mcpServers": {
+    "agentopus": {
+      "url": "https://api.opus.pro/api/agent-mcp"
+    }
+  }
+}
+```
+
+Or use the product's **Add to Cursor** deeplink from https://agent.opus.pro/mcp.
+Complete OAuth on first tool call.
+
+### Claude Code
+
+```bash
+claude mcp add --transport http agentopus https://api.opus.pro/api/agent-mcp
+```
+
+### Codex CLI
+
+```bash
+codex mcp add agentopus --url https://api.opus.pro/api/agent-mcp
+codex mcp login agentopus
+```
+
+## Tool surface (18 × `agentopus_*`)
+
+Client tools are prefixed `agentopus_`. Core flow:
+
+1. `whoami` — user, org, plan, credit quota (works on any plan; generation needs Pro+)
+2. `list_styles` / `list_voices` / `list_image_models` / `list_assets` — catalog
+3. `prepare_project` — prompt or script; voice, style, aspect ratio, duration, ≤8 anchors
+4. `start_project` — queue render (**consumes credits**)
+5. `get_video` — poll until result URL ready
+6. `list_projects` — status + result URLs
+
+Also: asset library (`upload_asset` / `register_asset` / `delete_asset`),
+one-off project uploads, `customize_style`, `clone_voice` / `delete_voice`.
+
+## Limits (product-stated)
+
+- Max video length ~**600s**
+- Bulk: **10** `start_project` calls per **5 minutes** per org (prepared projects kept)
+- Asset library ≤**10** saved anchors; ≤**8** per video
+- Image formats for anchors: png, jpeg, webp, gif
+- Voice clone: **10–30s** non-instrumental sample
+
+## HITL rules (KK)
+
+- Always `whoami` before a spendy batch.
+- Prefer a full script or detailed brief before `start_project` — AgentOpus
+  auto-confirms brief/transcript/storyboard after ~5 minutes of idle if you
+  walk away in the web app; treat MCP the same: review prepared intent.
+- Brand logos / palette: upload anchors; do not rely on prompt-only brand
+  description (product limitation).
+- After a keeper lands, stills/thumbs for the KB still go through **Rafiki**;
+  meetup short-form distribution still goes through **OpusClip / Buffer** in kk-kb.
+
+## Related
+
+- Local Rafiki MCP: [`MCP.md`](./MCP.md)
+- Floyo video: [`FLOYO.md`](./FLOYO.md)
+- kk-kb OpusClip clipping MCP: `kk-kb/tools/media_pipeline/OPUS-MCP.md`
+- kk-kb image shim: `kk-kb/tools/image-gen/README.md`

docs/FLOYO.md

@@ -4,6 +4,9 @@ Rafiki can drive [Floyo](https://flowyo.ai) — a hosted ComfyUI service — to
 video clips from a workflow template. This is the first slice of folding the
 [alex-samuel](PERSONAL-MEDIA-SUITE.md) film pipeline into Rafiki.
 
+Need a full narrative / explainer / promo video from a prompt or script instead?
+Use hosted **AgentOpus MCP** — see [`AGENT-OPUS-MCP.md`](./AGENT-OPUS-MCP.md).
+
 **Dry-run-first.** Nothing uploads, submits, or downloads unless you pass `--execute`. The
 only thing that spends FloTime is `--execute`.
 

docs/INDEX.md

@@ -25,6 +25,8 @@ deeper operating docs by surface area.
   and safety notes.
 - [MCP Output Contract](MCP-OUTPUT-CONTRACT.md) - proposed stable JSON envelope
   for typed tool outputs (paths, URLs, counts, errors, mutation flags).
+- [AgentOpus MCP](AGENT-OPUS-MCP.md) - hosted generative-video MCP (prompt/script
+  → finished video); same lane as Rafiki/Floyo, not OpusClip clipping.
 - [Floyo Video](FLOYO.md) - drive Floyo (flowyo.ai) hosted-ComfyUI workflows to
   render short clips; dry-run-first CLI + MCP.
 - [Doctor](DOCTOR.md) - what `npm run doctor` / `npx rafiki --doctor` checks,

docs/MCP.md

@@ -4,6 +4,11 @@ Rafiki ships a local MCP server at `mcp_server.py`. It exposes the common
 image-generation functions directly, typed wrappers for stable local workflows,
 and a constrained bridge to the canonical CLI for everything else.
 
+For **hosted narrative video generation** (prompt/script → finished video via
+Agent Opus), see [`AGENT-OPUS-MCP.md`](./AGENT-OPUS-MCP.md). That is a separate
+remote MCP — keep it next to Rafiki/Floyo in your mental model; do not confuse
+it with OpusClip long-form → shorts in kk-kb.
+
 The server uses the official Python MCP SDK's `FastMCP` API and runs over
 stdio by default, which is the right fit for local desktop tools.