Product
Cursor for marketing
Marketing product path
Feature map
The core workflow and what is being built next
Available paths, beta surfaces, and planned work are labeled so the product does not overclaim.
Brand memory
in appURL in. Voice, ICP, pillars out.
Onboard with a product or business URL. We extract ICP, tone, and pillars for SaaS, startups, and MSMEs so every draft stays on-voice.
Start onboarding →Agent fleet (AI CMO wall)
productNamed workflows. HITL control. Provider-confirmed publishing.
Named workflows with honest status: brand intake, social drafts, Reddit/HN/SEO (live); Writer/GEO/publish/analyst (partial); experimental agents later. Cursor-like HITL + saved brand context.
See fleet →Operating loop
productCONTEXT → PLAN → CREATE → APPROVE → PUBLISH → REVIEW
One focused loop from product context to approved campaigns and reviewable outcomes. Not a calendar. Not a chat box. Work moves through clear stages with human approval.
How it works →Studio
in appDrafts from brand context.
Generate a 7-day campaign brief plus channel-ready posts from memory — X, LinkedIn, Reddit. Edit in the HITL queue before live publish. L2 can auto-queue X/LinkedIn drafts only. Google Business Profile coming soon.
Open Studio →HITL approval queue
in appAutonomy with a human gate.
Set how far agents can go per channel. High-risk engages always stop. Approve or reject drafts before they hit the wire.
Open queue →Connectors
productOAuth when you need it.
Reddit, X, and LinkedIn via managed OAuth. Approve first — publish only after a real provider post ID. Google Business Profile coming soon (labeled).
Connect accounts →SEO / AEO
productRankings and LLM citations.
Structure, guides, sitemap, robots, and llms.txt so humans and models can find you. Distribution includes being citeable.
View llms.txt →Sandboxed execution
productUntrusted data never becomes a tool call.
Scraped pages, decks, and skill files are stored and cited as data. HITL gates every live post; generation and approve paths are rate-limited per workspace so one session cannot drain providers.
Security posture →
Operating loop
Context → plan → create → approve → publish → review
Approval is always available — autonomy is a dial, not a one-way switch.
Save product, ICP, voice, and proof
Shape a focused campaign from brand context
Draft posts, threads, assets
HITL quality and brand check
Provider-confirmed publishing
Review what shipped and what to change
Fleet
AI CMO agent wall
Category floor: named agents for the jobs a marketing team does. Our wedge: brand memory that does not reset, HITL like Cursor accept, publish only with real provider post IDs — not a fake “always live” CMO.
Brand + Memory
URL → durable brand memory (core / semantic / episodic).
Strategist
Owns goals, calendar, and channel mix from brand memory.
Writer + Social
Long-form and channel drafts (X · LinkedIn · Reddit) in your voice.
Reddit / HN
Opportunity loops: find threads, draft helpful replies for HITL.
SEO + GEO
Keyword drafts, rankings, and LLM citations (llms.txt / schema).
Distributor + HITL
Approve → publish only with real provider post IDs.
Analyst
Reviews published work; GSC/GA and attribution next.
UGC / Video (later)
Short-form briefs and clips are planned after the core loop is proven.
Full catalog · honest status (code-audited)
Live = end-to-end path in code (fails closed without keys). Partial = real path, incomplete vs definition of done. Soon/later = not built.
Brand intake
LivePaste a product URL → structured brand memory (ICP, tone, pillars, never-say).
Live: POST /api/brand — Firecrawl scrape + model extract + store + SuperMemory sync. Fails closed without Firecrawl/model.
Open →Memory keeper
LiveLayered core / semantic / episodic memory so context never resets.
Live: brand-memory layers + SuperMemory recall; approve/reject write episodes. Without SuperMemory, core brand fields still persist in Supabase.
Open →Strategist
LiveGoals, channel mix, and 7-day campaign brief from brand memory.
Live: POST /api/marketing/campaign — model + SuperMemory → 7-day brief stored. No static templates; fails closed without keys. Drafts still generated separately via Studio HITL.
Open →Social content
LiveChannel-ready short drafts (X, LinkedIn, Reddit) from memory.
Live: POST /api/marketing/draft — model + SuperMemory → 3 posts. L1 all pending; L2 auto-queues X/LinkedIn (not publish); Reddit stays pending. Fails closed without OpenAI/SuperMemory.
Open →Reddit agent
LiveFind relevant threads; draft helpful replies for review (never spam auto-post).
Live: runRedditAgent + POST /api/marketing/agents/reddit. Fail closed without OPENAI. Never auto-posts.
Open →Hacker News agent
LiveSurface moments to share; draft comments for founder review.
Live: runHnAgent + POST /api/marketing/agents/hn. Fail closed without OPENAI. Never auto-posts.
Open →SEO agent
LiveKeyword gaps, outlines, and blog drafts for approval.
Live: runSeoAgent + optional Tavily research. Fail closed without OPENAI. Long-form stored as queue post (platform linkedin for body length).
Open →HITL autonomy
LiveL1–L3 dial; human gate on brand-risk actions.
Live: queue edit/approve/reject + L1–L3 dial. L2 auto-queues low-risk social drafts only. L3 blocked. Generation + approve rate-limited per workspace.
Open →Writer
PartialLong-form articles and landing copy in brand voice.
Partial: long-form is the SEO agent (blog draft → queue as LinkedIn-platform long text). No separate landing-page writer product surface yet.
Open →X agent
PartialDrafts and threads; publish only via connected account + provider post ID.
Partial: drafts via social content; OAuth connect via Composio. Live publish only when account ACTIVE and provider returns post ID — not a dedicated “X agent” run button.
Open →LinkedIn agent
PartialProfessional posts and ideas; HITL then connected publish.
Partial: same as X — draft path live; publish depends on Composio connection + provider confirm.
Open →GEO / AEO agent
PartialMake the brand citeable by LLMs (structure, llms.txt, FAQ, schema).
Partial: heuristic site scorecard (POST /api/marketing/site-audit) + recommendations. Does not auto-generate or deploy llms.txt/FAQ/schema onto the customer’s domain. vibemarketer’s own /llms.txt is separate.
Open →Distributor
PartialExecute approved posts through OAuth connectors; never invent post IDs.
Partial: approve → queued; confirmPublished only with real provider id (no synthetic ids). Live publish depends on Composio ACTIVE accounts. Per-owner publish rate limit on approve.
Open →Analyst
PartialReport what shipped and what worked; feed LEARN.
Partial: weekly rollup from marketing_state counts. No GSC/GA metrics until OAuth + tool pull is live.
Open →UGC / video
PartialShort-form briefs and clips for social (Remotion / templates).
Partial: promo-video package + optional Grok Imagine creative on a draft. Not a full UGC agent fleet.
No app surface yet
Influencer
LaterDiscover creators and draft outreach (always human-sent).
Not built — catalog placeholder only.
No app surface yet
Coding / tech SEO
LaterSuggest technical SEO fixes (meta, schema) as reviewable diffs.
Not built — catalog placeholder only.
No app surface yet
Link broker
LaterBacklink opportunities — high risk; last among growth agents.
Not built — deferred on purpose.
No app surface yet
Machine-readable catalog: /api/marketing/agents. Competitive research: Okara-style AI CMO is the checklist; vibemarketer is the IDE + runtime.
Trust
Security & control
Autonomy dial
Draft-only, approve-to-publish, or supervised engage — per channel. High-risk actions always hit the HITL queue.
Untrusted ingest
Scraped pages, decks, and skill files are data — stored and cited, never allowed to issue tool calls. Tokens stay server-side.
Start free
URL → brand → drafts → approve
One product promise. Create an account and paste your site.
