SocialSeed.ing By Agent — 26 Preview Mockups

Persona: Technical solo founder / indie DTC operator who already lives in a terminal (deploys with a Makefile, configures with .toml/.env) and distrusts another login-walled SaaS dashboard for something as mission-critical as ad spend

Killer feature: Campaign-as-code: your whole seeding strategy is a diffable `campaign.toml` (budget cap, target CPV, creator filters, #ad policy) — `git commit` it, `seedctl hire` it, and the agent's every reallocation decision lands as an append-only line in `~/.seedctl/ledger.jsonl` you can audit, grep, and replay. No black-box dashboard between you and your money.

Key screens: $ seedctl hire ./campaign.toml — boot output: ADK sub-agents (strategy/sourcing/briefs/publish/measure/reallocate) coming online over A2A · campaign.toml — the editable plain-text campaign spec (budget_cap, target_cpv, creator filters, mandatory #ad disclosure block) · seedctl console — curses TUI: live verified-view counter, spend-vs-cap gauge, agent action log, fraud-filter rejects ticker · seedctl bill --format=table — the $0.01/verified-view ledger, fraud-filtered, with `| jq` JSON output shown alongside · seedctl agents — A2A sub-agent status board (PID, last decision, MCP tools bound: mock-tiktok, metrics-sim) · seedctl kill <creator_id> / seedctl scale — manual override of the reallocation agent from the prompt

Why this lens: Every other advocate will mock a polished web app; the contrarian take is that the people who'd actually trust an autonomous agent with a live budget are terminal-native operators who want campaign-as-code, plain-text auditability, Unix-pipeable billing, and zero vendor dashboard lock-in. Same ADK/MCP/A2A/Gemini guts — radically unfashionable surface.

↗ open full mockup

Persona: The DTC founder who is now the de-facto on-call operator for an autonomous agent spending real money on their behalf — wants a status board, alerts, and an audit trail before they trust it overnight

Killer feature: The Verified-View Ledger + Budget Circuit Breaker: every cent is line-itemed to a creator/post with its fraud-filter verdict, a burn-rate gauge projects when the cap trips, and a one-click 'PAUSE ALL' kill switch with an audit log of who/what/when.

Key screens: Campaign Health Board — grid of status cards: Verified Views (rate/min), Fraud-Filter Pass %, Budget Burn vs Cap, Creator Pipeline Stage, Publish Queue depth, Sub-agent A2A latency — green/amber/red like an uptime board · Verified-View Ledger — append-only table: timestamp, creator, post, raw views, dedup/watch-time/source-anomaly/attribution verdicts, billable views, $ charged, running total vs deposit · On-Call Feed — paging-style alert stream: 'Fraud spike on @creator', 'Burn rate → cap in 3h', 'Publish failed (mock TikTok 503) — retry 2/3', each with ack / snooze / escalate · Agent Runbooks — one card per sub-agent (Strategy, Sourcing, Brief, Publish, Measure, Reallocate): last run, success rate, owned MCP tools, and the documented failure-mode playbook (e.g. 'Brief missing #ad disclosure → block publish, regenerate') · Budget Circuit Breaker — burn-rate gauge, projected trip time, hard-cap deposit bar, and the big PAUSE ALL / RESUME control with confirmation + reason log

Why this lens: Every other advocate sells the magic of hiring the agent; I sell the day-2 reality — you are now operating an autonomous spender, so the product surface is a health board, a to-the-cent ledger, alerts, and runbooks, not a glossy 'set it and forget it' pitch.

↗ open full mockup

Persona: Solo DTC founder on TikTok Shop who wants a campaign running before lunch, not a vendor call

Killer feature: 60-second Hire flow: zero onboarding screens — product URL + budget cap is the entire setup; the ADK agent swarm boots, drafts strategy, and starts creator outreach while you watch the meter tick

Key screens: Hero one-liner + single input (product URL) + budget slider + giant 'Hire the Agent' CTA — no nav, no pricing page · 60-second boot sequence: live log of sub-agents waking up (Strategy → Creator Sourcing → Brief Writer → Publisher) with #ad disclosure auto-stamped · Live campaign console: verified-views counter, spend-vs-cap bar, creators in-flight, GMV-to-date — single viewport · Instant billing readout: $0.01 × verified views = spend, fraud-filtered count vs raw count, hard-cap auto-stop status · One-tap 'Top up budget' / 'Pause agent' controls inline on the console

Why this lens: Every other take builds a dashboard you configure; this one collapses setup to a URL + a slider and gets a campaign live in under a minute — speed-to-launch is the product. No projects, no workspaces, no creator marketplace browsing.

↗ open full mockup

Persona: Bootstrapped solo DTC founder on TikTok Shop who runs the whole business from a laptop, has ~$300–$2k of growth budget per launch, and refuses to add another paid SaaS line item

Killer feature: Cent-level live ledger with a hard budget cap: the ADK agent debits your deposit one verified view at a time, shows fraud-rejected views as $0 (you never pay for them), and auto-halts at exactly your deposit — no overage, no surprise invoice, no monthly fee.

Key screens: Deposit & cap screen — enter budget (e.g. $500 = 50,000 verified views), pick product from TikTok Shop mock; one number, no plan tiers · Live ledger — append-only table: timestamp, creator, raw views, verified views, fraud-rejected (shown $0), running spend, remaining cents, projected exhaustion date · Cost-per-outcome panel — actual CPV vs target $0.01, cost per TikTok Shop click, cost per attributed order; big numbers, low-fi · Agent activity log — A2A sub-agent events (strategy → sourcing → brief → publish → measure → reallocate) as plain text lines, each tagged with infra cost ($0.00 on free tier) · Campaign close report — one-page export (static HTML, no PDF service): total verified views, GMV attributed, leftover deposit auto-refunded, screenshot-ready

Why this lens: Everyone else sells a polished campaign cockpit; this is a bare wireframe whose entire reason to exist is proving the unit economics — $0/mo to operate, $0.01/view pay-as-you-go, fraud views cost you literally nothing, and the ledger is append-only so the brand can audit the agent like a bank statement.

↗ open full mockup

Persona: Solo founder of a small DTC brand on TikTok Shop who has zero hours for influencer ops and wants the feeling of a marketing team without the headcount

Killer feature: The 'Hire' moment: a single elegant decision (product + budget + tone) replaces weeks of creator-ops setup, and from then on the agent works in the open with a console that feels like watching a great hire do great work.

Key screens: Hire the Agent — hero: product card + budget dial + brand-voice chips, one big 'Hire Mira' button, animated empty state showing what the first 48h will look like · Onboarding handshake — the agent introduces its sub-team (Strategy, Sourcing, Briefs, Publishing, Measurement) and the rules it will never break (#ad disclosure, fraud filter, hard budget cap) · Live campaign console — calm timeline of agent actions, a 'budget burndown' ring (verified views x $0.01 vs deposit), creators going live, kill-and-scale moves explained in plain language · Verified view detail — tap any view batch to see the fraud filter pass (dedup / watch-time / source / attribution) so the spend feels earned · Billing report — a delightful end-of-campaign recap: CPV achieved, GMV attributed, setup-to-launch time, one-tap re-hire for the next launch

Why this lens: Every other take builds a control panel; I build a relationship. The interface treats the agent as a trustworthy hire — emotional confidence and radical transparency over knobs and tables.

↗ open full mockup

Persona: Hands-on growth lead / founder-operator at a small TikTok Shop DTC brand who lives in spreadsheets and wants every creator, post, view, and dollar in one sortable grid — not a glossy dashboard

Killer feature: A live per-post metrics grid where every column (verified views, fraud-rejected views, watch-time %, CTR→Shop, GMV, CPV, $ spent) is sortable and filterable, and a Budget-Burn ledger row that hard-stops the agent when verified-views × $0.01 hits the deposit

Key screens: Creator Roster grid — every sourced micro-creator as a row: handle, niche, followers, avg views, fit score, A2A status (sourcing→briefed→posted→measured), #ad-compliance flag, $ allocated · Post Metrics grid — one row per published post: creator, post URL, posted-at, raw views, verified views, fraud-rejected (dedup/watch-time/anomaly/attribution breakdown on hover), CTR→TikTok Shop, orders, GMV, CPV, $ billed — fully sortable column headers · Budget-Burn ledger — chronological spend rows: timestamp, event (view batch verified / creator paid sim / reallocation), delta $, running total, % of cap, projected stop-date; red banner at cap · Reallocation log grid — agent kill-and-scale decisions as rows: timestamp, action (kill creator X / scale creator Y +$), trigger metric, before/after CPV — the autonomy audit trail · Campaign setup row-form — product, deposit ($), target CPV, niches, region, disclosure preset → one click spawns the ADK agent run and seeds the grids

Why this lens: Where the other previews show narrative cards, chat, or hero visuals, this is unapologetically a data grid: maximum information density, keyboard-first, sort/filter/group everything — it treats the autonomous agent as a worker whose output you audit in a table, not a chatbot you converse with

↗ open full mockup

Persona: Solo DTC founder running a TikTok Shop from their phone between meetings, school runs and warehouse trips — never opens a laptop dashboard, lives in notifications

Killer feature: One-thumb hire: tap 'Hire the agent', set a budget with a slider, confirm — the agent sources creators, writes #ad briefs, publishes via mock TikTok, and you only ever pay $0.01 per verified view. No dashboard, no laptop.

Key screens: Home — live campaign feed: vertical scroll of agent actions (sourced creator, brief sent, post live, view milestone), pull-to-refresh · Hire sheet — bottom sheet: pick product, drag budget slider ($→verified views estimate), toggle TikTok Shop link, big 'Hire the agent' CTA · Creator cards — full-bleed swipeable cards (TikTok-style): creator stats + agent's reasoning; swipe right to approve, left to skip, agent auto-fills the rest · Budget alert — push-style banner + sheet: 'You've used 80% of your $500 deposit (39,800 verified views)'; tap to top up or let it stop · Billing card — thumb-friendly card: verified views, fraud-filtered views removed, current spend vs cap, 'Add deposit' button · Campaign report — swipe-up summary: CPV ≈ $0.01, clicks to Shop, GMV attributed, setup-to-launch time

Why this lens: Every other preview builds a dashboard or console; P07 assumes the buyer never opens a laptop. The entire hire→approve→monitor→pay loop fits in a phone, runs on notifications, and uses TikTok's own UI grammar (full-bleed swipe cards, bottom sheets) so a founder feels at home.

↗ open full mockup

Persona:

Killer feature: The whole campaign lifecycle is conversational ops: every agent decision (strategy, creator picks, kill-and-scale, spend) arrives as a Slack card you can approve or veto with a button — no dashboard tab to keep open, no login, the agent just pings the channel.

Key screens: Slash command palette: typing `/seed hire` shows args (brand handle, deposit, product URL) and a confirm modal · Agent intro thread: 'Hired ✅ — deposit $300 = 30,000 verified views cap' + a posted Strategy card (audience, hooks, #ad disclosure note) · Creator-sourcing card: 5 micro-creators with fit score / est. CPV / niche, Approve all · Swap · Reject buttons · Daily campaign update message: views, verified %, fraud-filtered count, GMV→TikTok Shop clicks, spend bar against cap · Budget-cap warning: '⚠️ 92% of $300 deposit consumed — agent will pause posting at 100%. Top up?' with a Top up $100 button · Final report card in-thread: 31,204 verified views × $0.01 = $312 (capped at $300), CPV $0.0096, orders attributed, link to full web report

Why this lens: Every other advocate builds a web dashboard or a hero landing page; P08 treats Slack as the product surface — the agent is a teammate in a channel, billing receipts and kill/scale decisions are messages, and the web app is demoted to a passive report archive.

↗ open full mockup

Persona: Technical founder / growth engineer at a small TikTok Shop DTC brand who already lives in a terminal, ships via CI, and wants campaigns as code — not another SaaS dashboard tab

Killer feature: Campaigns-as-code: a versioned brief.yaml + a single `hire` command spins up the ADK agent swarm (strategist → sourcer → briefer → publisher → measurer → reallocator), streams structured JSONL agent logs to stdout, enforces the hard budget cap as a process that exits non-zero when deposit is exhausted, and `socialseed bill --format csv` emits a per-view audit ledger with the fraud-filter verdict on every view.

Key screens: `socialseed hire --budget 300 --product ./brief.yaml --dry-run` — agent plan preview: proposed strategy, creator shortlist, projected verified views @ $0.01, #ad disclosure check, budget cap math; exit 0 if plan valid · `socialseed campaign watch cmp_8f2a` — live streaming TUI: A2A message timeline (strategist→sourcer→briefer→publisher→measurer), running verified-view counter, spend bar vs cap, fraud-filter rejects ticking by · `socialseed bill cmp_8f2a --format table` — ledger: each verified view with watch_time, dedup_hash, source_score, attribution, ✓/✗ verdict, line cost $0.01, running total, hard-cap line; `--format csv` pipes clean · `socialseed creators ls cmp_8f2a` — sourced micro-creators table: handle, niche, followers, projected views, brief status, post URL (mock), disclosure=#ad ✓ · `socialseed campaign reallocate cmp_8f2a --auto` — reallocator agent output: kills underperformers, scales winners, diff of budget split before/after, remaining cap · `socialseed report cmp_8f2a` — end-of-campaign ROI: verified views, actual CPV, TikTok Shop click-throughs, attributed GMV/orders, setup→launch time; `--json` for CI dashboards

Why this lens: Every other preview is a web dashboard. This one treats the agent as a unix process you invoke, observe, pipe, and exit-code-check — campaigns are YAML in your repo, agent reasoning is JSONL on stdout, billing is a CSV you can diff. Built for the founder who'd rather `cron` a campaign than click one.

↗ open full mockup

Persona: By 2031, the DTC founder who never opens a marketing tool: they post a budget and an outcome to an open exchange and hire whichever autonomous growth-agent is winning on verified CPV that week.

Killer feature: An open marketplace where brand-agents and creator-agents auto-negotiate over A2A in real time, settling on a guaranteed $0.01/verified-view price with a programmatic budget cap and fraud oracle baked into the contract — the agent that under-delivers loses the next auction automatically.

Key screens: The Exchange floor — a real-time map of open brand briefs, agent bids streaming in, and a 'who's winning on CPV' leaderboard of hireable growth-agents · Hire flow — drop a product link + a deposit, watch three rival agents pitch a strategy/creator-mix/projected-views in seconds, tap one to sign the performance contract · Live negotiation theater — your hired brand-agent talking to creator-agents over A2A: offers, counteroffers, brief drafts with auto-inserted #ad disclosure, deals closing live · GMV stream — verified-views ticker climbing toward the budget cap, fraud-filter rejections shown inline, TikTok Shop click→order attribution flowing as it happens · Auto-reallocation board — agent killing under-performing creators and scaling winners mid-flight, each move logged as an explainable decision · Settlement & reputation — campaign close-out: actual CPV vs $0.01 target, GMV report, and the agent's updated trust score that decides its next bid

Why this lens: Every other preview ships a dashboard or a single agent you operate. This one is a two-sided marketplace: brands don't manage anything, they buy outcomes on an open exchange where autonomous agents compete, negotiate agent-to-agent, and are paid only for fraud-verified results — Upwork for marketing AIs, with the contract enforced in code.

↗ open full mockup

Persona: Solo founder of a small DTC brand on TikTok Shop who wants a campaign run for them and just needs to see status + what they've been billed

Killer feature: A single 'Hire' form that turns product URL + budget into a running campaign, plus a billing report that proves you paid exactly $0.01 per verified view (fraud-filtered views excluded, hard cap enforced).

Key screens: Hire the Agent — one form: product URL, budget deposit, target audience note, #ad disclosure checkbox (pre-checked, locked) · Campaign Status — verified views counter, budget burned vs cap, creators sourced, posts live, next agent action · Activity Log — plain reverse-chronological list of what each sub-agent did (strategy, sourcing, brief, publish, measure, reallocate) · Billing Report — line items: verified views x $0.01, fraud-filtered views excluded (count shown), running total vs deposit, CSV export · Campaigns list — table of past/active campaigns with CPV, GMV attributed, status

Why this lens: No dashboard theater, no charts-for-charts'-sake, no new framework. The minimum surface a busy founder actually needs: hire, watch, get billed. Built on a 3-year-old stack so it ships this week.

↗ open full mockup

Persona: Privacy/compliance-conscious DTC founder (and the creators they touch) who wants TikTok seeding outcomes without their product roadmap, customer data, or creators' personal info leaving a sealed boundary

Killer feature: Privacy-by-default autonomy: every sub-agent runs against tokenized creator handles + a need-to-know data scope, view verification uses k-anonymous aggregate watch-time buckets (no per-user tracking, no device graph), and an immutable audit log lets the brand prove — and the creator confirm — exactly what the agent saw and did.

Key screens: Sealed Campaign overview — the boundary diagram: what's inside the vault (brief, budget, GMV signals), what's outside (no TikTok personal data ingestion), green 'data did not leave' status badge · Consent & Capability panel — per sub-agent (Strategy / Sourcing / Brief / Publish / Measure / Reallocate) a CAN-SEE vs CANNOT-SEE table; toggle to revoke a scope mid-campaign · Creator PII minimization view — creators shown as tokenized handles + coarse buckets (region tier, audience-size band); raw handles only revealed at the moment of human hand-off, logged · Audit log — append-only, hash-chained list of every agent action (sourced creator #c8f2, sent brief v3 with #ad clause, paused spend at cap), exportable, with 'verify chain' button · Verified-view ledger — billing from k-anonymized aggregate watch-time buckets; fraud filter shown as cohort-level anomaly flags, never individual profiles; hard budget cap meter · Data Subject / Creator portal — a creator opens a link, sees exactly which brand campaign holds a token for them, what brief they were sent, and a 'forget me' button that purges the token

Why this lens: Where the other previews show dashboards, GMV charts, or creator marketplaces, this one treats the campaign as a sealed enclave: data minimization, tokenized creators, k-anonymous view counting, revocable per-agent scopes, and a hash-chained audit log are the product — autonomy you can prove was contained.

↗ open full mockup

Persona: Analytically-minded growth lead at a small DTC TikTok Shop brand who wants to watch verified-view velocity, CPV distribution, and GMV attribution in real time rather than read a weekly recap

Killer feature: Live verified-view ledger with a fraud-filter funnel: watch raw views → dedup → watch-time threshold → source-anomaly → attribution, with the billable count and running spend-vs-cap updating per event

Key screens: Campaign overview KPI strip — verified views, billable spend / budget cap, actual CPV, attributed GMV, ROAS — with sparklines · View-velocity time series — stacked area of verified views by creator cohort, with reallocation events annotated on the curve · Fraud-filter funnel — raw events to billable views, each stage's drop %, anomaly sources broken out · Creator-cohort performance grid — micro vs mid creators: views, watch-time, CPV, click-through, GMV-per-view, agent kill/scale flags · CPV distribution histogram — per-view cost spread vs the $0.01 target line, with budget-cap burn-down · Attribution waterfall — verified views → TikTok Shop clicks → add-to-cart → orders → GMV, each step's conversion rate

Why this lens: Where other previews sell the autonomy or the chat, this one IS the measurement layer — a quant cockpit that makes performance-priced billing legible: you can audit the fraud filter and the GMV attribution chain, not just trust it

↗ open full mockup

Persona: First-time TikTok-Shop DTC founder who has never run creator marketing and wants to actually understand what the agent is doing on their behalf — not just trust a black box

Killer feature: The 'Why this?' panel on every autonomous decision: the agent shows its reasoning, the data it used, the rule it obeyed (budget cap, fraud filter, #ad disclosure), and a one-line lesson — turning each campaign into a TikTok-marketing course you didn't pay extra for

Key screens: Onboarding walkthrough — 6-step interactive tour of the campaign loop with a live progress bar; each step explains what that sub-agent does and why it matters before you ever spend a cent · Strategy card with 'Why this?' — agent proposes audience, hook angles, creator tier mix and budget split; expandable explainer cites the synthetic-data signals it used + a 'Lesson: why micro-creators beat one mega-post' note · Creator sourcing board — shortlisted mock creators with fit scores; hover tooltips define engagement rate, fake-follower flags, niche overlap; sidebar mini-lesson on vetting creators · AI brief preview — generated creator brief showing the mandatory #ad / branded-content disclosure block highlighted, with a tooltip explaining the FTC/TikTok rule and why the agent cannot remove it · Live campaign loop dashboard — verified-views counter ticking toward the hard budget cap, fraud-filter rejections shown with reasons, kill-and-scale reallocations narrated step by step · ROI report + 'What you learned' recap — CPV vs target, GMV attributed, plus a generated summary of the 5 creator-marketing concepts this campaign just taught the founder

Why this lens: Every other preview hides the agent's reasoning behind a slick dashboard; this one is pedagogically opinionated — the agent is autonomous AND a teacher, so the founder ends each campaign able to sanity-check the next one. Compliance rules (#ad, fraud filter, budget cap) are reframed as teaching moments instead of fine print.

↗ open full mockup

Persona: Two-sided from day one: (1) DTC growth leads who hire the agent for a campaign, and (2) TikTok micro-creators who join the shared pool to get matched, briefed, and paid — plus agency teams running multiple brand seats inside one workspace.

Killer feature: Network-effect creator pool: the agent sources from a community of opt-in creators with cross-campaign verified-view track records and ratings, so each campaign that runs makes matching faster and CPV tighter for everyone — the pool is the moat, not the model.

Key screens: Campaigns Marketplace feed — agent-posted open briefs (budget cap, target views, #ad terms), creators apply, brands watch fill-rate in real time · Creator Pool directory — rated creator cards (verified views delivered, watch-time quality score, niche, prior brands), agent-recommended shortlists · Shared campaign room — brand seat, accepted creators, and the orchestrator agent in one thread with @mentions, brief approvals, and #ad disclosure checks · Creator profile + reputation — lifetime verified views, fraud-flag history, ratings from brands and the agent's measurement sub-agent · Agency multi-seat console — switch between brand workspaces, pooled creator favorites, team roles (owner / strategist / viewer) · Live campaign ledger — verified-views counter vs budget cap, per-creator contribution, kill-and-scale reallocation log visible to all participants

Why this lens: Every other take ships a single-tenant agent + ROI dashboard for one brand. SeedHive is multi-user day one: brands and creators inhabit the same workspace, a marketplace feed, and a reputation graph — the agent gets smarter and cheaper per view as the network grows, which no solo-tenant product can replicate.

↗ open full mockup

Persona: A one-person DTC brand owner — does product, support, shipping and growth alone — who wants TikTok views without hiring anyone, learning influencer ops, or babysitting a dashboard

Killer feature: Zero-ops autonomy: a single 'Hire Agent' button spins up the full ADK multi-agent crew (strategy → sourcing → briefs → publish → measure → reallocate); a hard budget cap means it physically cannot overspend your deposit while you sleep.

Key screens: Command Center (default) — agent status pill (Working / Paused / Out of budget), live verified-view counter, deposit burn-down bar, one big Hire/Pause button · Hire Agent sheet — paste product URL + monthly cap + 1-line vibe, that's the entire setup; agent launches in minutes · Top-up sheet — add to deposit, set an auto-topup threshold so it never stalls overnight · The Bill — verified views × $0.01, fraud-filtered views excluded (shown as a line item), GMV attributed, downloadable receipt · Activity log — append-only feed of what the agent did (creators booked, briefs sent w/ #ad enforced, posts published, budget reallocated) — read-only, no approvals needed · Settings — notification minimalism: one weekly digest email + a 'ping me only if out of budget' toggle

Why this lens: Every other advocate builds a workspace you operate. I build a hire — one page, one button, no team, no approvals, no ops. The agent is staff, not software; the UI is a payroll stub plus an off switch.

↗ open full mockup

Persona: VP of Procurement / Head of Brand Marketing Ops at a multi-brand CPG holding company evaluating an autonomous TikTok-seeding agent for org-wide rollout via Google Cloud Marketplace

Killer feature: Contractual, metered autonomy: every verified view is a line item in an immutable audit trail with an SLA (fraud-filter SLA, hard-cap SLA, disclosure-compliance SLA) — exportable as a procurement-grade invoice and consumed against your committed Google Cloud spend

Key screens: SSO & Access — SAML/OIDC config, SCIM user provisioning, role matrix (Org Admin / Brand Manager / Finance Viewer / Auditor / Read-only) · Portfolio Rollup — all brands under the holding company, per-brand verified-view spend vs. cap, blended CPV, GMV attributed, compliance status · Audit Trail — append-only event log: agent actions, brief approvals, budget-cap halts, fraud-filter rejections, disclosure checks, with actor / timestamp / hash · Marketplace & Gemini Enterprise — Google Cloud Marketplace listing card, private offer terms, Agent Engine deployment region, Gemini Enterprise connector status, committed-use drawdown · SLA & Metering Monitor — live $0.01/verified-view meter, fraud-filter accuracy SLA, budget-cap enforcement SLA, uptime SLA, credits ledger for breaches · Billing Exports — month-close invoice per cost center, CSV/JSON/PDF export, PO matching, tax fields, verified-view reconciliation file

Why this lens: Where the other previews sell the agent's marketing magic, this one sells the contract around it: identity federation, least-privilege RBAC, an immutable audit log, written SLAs on the metering and fraud filter, and a one-click Google Cloud Marketplace procurement path — the artifacts that turn a cool demo into an enterprise PO.

↗ open full mockup

Persona: Solo founder of a small DTC brand on TikTok Shop who wants to *watch* an autonomous agent run their seeding campaign like a story unfolding, not babysit a dashboard

Killer feature: The Campaign Recap: a single auto-generated, screenshot-ready story card (views delivered, real CPV, GMV, top creators, budget used vs cap) the founder can post or send to investors — the agent's work made into a narrative artifact

Key screens: Hero / Press Play — bold type, the product + budget become a 'campaign film' you start; the agent's plan scrolls in as Act I · Act II: Creators Cast — micro-creator cards animate in one by one with fit-score, #ad disclosure badge, and the AI brief they were handed · Act III: Live Feed — vertical timeline of publish pulses, verified-view ticks, fraud-filtered views struck through, GMV/click events surfacing · The Budget Gauge — a persistent draining meter (deposit ⇒ verified views × $0.01) that hits a hard stop and freezes the story · Kill & Scale Beat — a moment where the agent visibly reallocates: a creator card dims/exits, another brightens and doubles · Campaign Recap Card — the shareable closing scene: headline stats, CPV ≈ $0.01, top creators, before/after launch-time, #ad-compliant footnote

Why this lens: Every other preview is a control panel; this is a *narrative artifact* — the agent's autonomy is felt as a story you watch and can re-share, turning a B2B ops tool into something with emotional and social weight (matches the JTBD's 'look like a brand that gets TikTok')

↗ open full mockup

Persona: Platform engineers at Shopify-app vendors, headless-commerce platforms, and DTC OS startups who want to offer 'autonomous TikTok seeding, $0.01/verified view' as a feature inside their own product — never logging into a SocialSeed UI

Killer feature: client.campaigns.hire({ budget, product }) → an A2A-addressable autonomous campaign with a streaming verified-view ledger webhook and a hard budget cap enforced server-side — embeddable in any commerce stack without a single screen of our UI.

Key screens: Install & quickstart: npm i @socialseed/agent, client init, client.campaigns.hire({ budget, product }) returning a campaign handle · Webhook payloads: campaign.view.verified (with fraud-filter verdict, watch-time, dedup id, attribution) and campaign.budget.exhausted events · A2A agent card: socialseed-campaign-runner skills (strategy, creator-sourcing, brief-gen, publishing, measure-reallocate), endpoints, auth · MCP tool manifest: tools exposed to the host's own agent (campaigns.hire, campaigns.status, creators.search, ledger.read) · Billing & usage endpoint: GET /v1/campaigns/{id}/ledger → line items at $0.01/verified view, running total vs deposit cap · Example gallery: Shopify embedded app, headless Next.js storefront, internal Slack ops bot — all calling the same SDK

Why this lens: Every other preview ships a dashboard or chat UI; P19 ships zero end-user UI. The product IS the SDK + A2A endpoint + MCP manifest. Host apps (Shopify-app vendors, commerce platforms) embed it; SocialSeed never owns the customer relationship or the screen.

↗ open full mockup

Persona: Technical founder / growth engineer at a small DTC brand who would rather run an open-source agent on their own Vertex project than rent a black-box SaaS dashboard

Killer feature: The whole campaign autonomy stack (strategy → creator sourcing → AI briefs → publish → measure → reallocate, with fraud filter + hard budget cap) is MIT-licensed and runs on your own GCP project — no vendor lock-in, auditable billing math, governed in the open.

Key screens: README hero: badges (build, license MIT, ADK, Vertex AI, A2A, discord), one-paragraph what-it-is, `git clone` + `docker compose up` block · Agent graph: ADK orchestrator + sub-agents (StrategyAgent, SourcingAgent, BriefAgent, PublishAgent, MeasureAgent, ReallocateAgent) wired via A2A, MCP tool servers (mock-tiktok, metrics-sim, billing) shown as boxes · config.yaml panel: bring-your-own `vertex.project_id`, `region`, `gemini_model`, `budget_cap_usd`, `cpv_usd: 0.01`, `fraud_filter` thresholds · Transparent fee model: core agent = $0 (community-owned); optional hosted runner = $0.01/verified view pass-through + small ops margin, with the verification rules linked to source · CONTRIBUTING + open governance: how to add a creator-sourcing connector via MCP, RFC process, maintainers, code of conduct, 'no single company owns main' note · Issues / roadmap board: 'good first issue' labels, milestone for live TikTok API behind feature flag, fraud-filter test fixtures wanted

Why this lens: Every other advocate sells a hosted product or dashboard; this one ships the agent itself as a community-owned OSS project — the README is the product, GitHub is the funnel, and the pay-per-view billing is a removable add-on, not the moat.

↗ open full mockup

Persona: Bootstrapped DTC founder doing $5k–$50k/mo on TikTok Shop, no marketing hire, drowning in creator DMs and spreadsheets — wants growth on autopilot with a credit card, not a contract

Killer feature: Pure performance pricing wired straight to Stripe: prepay a budget, watch verified views tick up live, get billed per-view with a fraud filter, hard budget cap, and instant refund of the remainder — zero retainer, zero risk.

Key screens: Hero: 'Hire a TikTok growth agent. Pay 1¢ per view.' + fat 'Start with $100' CTA · Live counter: 'X agents working right now · Y verified views delivered today' · How it works in 3 steps: drop your TikTok Shop link → set a budget → agent sources creators, briefs, publishes, reallocates · Transparent pricing block: $0.01/verified view, pay-as-you-grow tiers (Starter free / Pro $29 / Scale $99) on top, hard cap + refund guarantee · Social proof strip: founder logos, '2,300 verified views per $23', testimonials · Billing & receipts panel: current spend vs deposit, next invoice, cancel/refund button

Why this lens: Everyone else ships a campaign dashboard; I ship a checkout page. The product IS the pricing — 1¢/verified-view, prepaid, capped, refundable — packaged as a launch-this-week SaaS aimed at $10k MRR, not an enterprise pilot.

↗ open full mockup

Persona: DTC growth lead who wants creator marketing run as a controlled experiment, not a vibes-based budget burn — willing to read a methods section

Killer feature: Pre-registered A/B/n hook-arm trials with sequential significance testing: the agent reallocates budget the moment a hook arm's CPV credible interval separates, and every kill/scale decision is logged with the test statistic that triggered it

Key screens: Experiment Design — pre-registration form: hypothesis (e.g. 'problem-first hook beats demo-first on CPV'), arms, sample size / power, stopping rule, budget cap as the hard horizon · Live Trial Board — A/B/n hook arms with running verified-view counts, CPV point estimate + 90% credible interval per arm, posterior-probability-best, and the auto-reallocation ledger · Methodology Panel — the fraud-verification protocol written as a methods section: dedup logic, watch-time threshold, source-anomaly z-score, attribution window — each with the synthetic-data validation it passed · Run Log — append-only reproducible event stream: every agent decision (creator sourced, brief generated with #ad clause, post scheduled, arm killed) with timestamp, inputs hash, and the ADK sub-agent that emitted it · Results Report — published 'What We Learned': effect sizes with CIs, winning hook characteristics, verified CPV vs target $0.01, GMV attributed, limitations & threats to validity, BibTeX-style citation block for the run

Why this lens: Other advocates show a dashboard or an autopilot; this is a research instrument — pre-registration, randomized arm assignment, sequential testing, explicit threats-to-validity, and a citable reproducible report. The agent does causal inference, so 'it worked' comes with a test statistic and a confidence interval, not a screenshot.

↗ open full mockup

Persona: Solo DTC founder on TikTok Shop who hates dashboards, loves seeing progress bars fill, and wants growth to feel like a game they're winning rather than a job they're losing

Killer feature: Performance-priced autonomy reframed as a game loop: $0.01/verified-view is the 'score', the hard budget cap is your 'energy bar', the agent's reallocation decisions show up as live 'manager moves' in the event ticker — real ops, weirdly addictive surface.

Key screens: Hire screen — pick your manager-agent personality (Hype Beast / Steady Hand / Bargain Hunter), set budget = starting energy, drop in product link · Campaign level (the main play screen) — manager character with speech bubble, big XP bar of verified views, energy/budget meter, creator unit roster with favorite stars, live event ticker ('Killed @dogfit_amy — CPV too high · Scaled @ploggers ⚡') · Creator unit card — a creator as a 'unit': stats (followers, watch-time, fraud-clean badge), the AI brief it was given (with mandatory #ad line highlighted), favorite toggle so the agent prioritizes them next time · Manager move log — every autonomous decision as a timestamped 'move' with the why (fraud filter caught bot views, budget cap throttle engaged, reallocated spend to top quartile) · Win screen — '30,240 verified views delivered! ⭐⭐⭐', CPV $0.0099, GMV attributed, energy spent vs cap, 'Run it back' button to clone the campaign · Trophy shelf — past campaigns as badges (First 100K Views, Under-Budget Hero, Zero-Fraud Run)

Why this lens: Every other advocate ships a dashboard or a console; P23 ships a management-sim game skin over the exact same ADK multi-agent ops — progress bars, units, energy, win screens — so the brutally B2B 'pay per verified view' loop becomes something a tired founder actually wants to open.

↗ open full mockup

Persona: The skeptical DTC founder who has been burned by 'autonomous' growth tools before: will not hand a credit card and a brand voice to a black box, wants to approve every creator, every brief, and every dollar — and wants a one-click exit the moment it feels wrong

Killer feature: Approval Mode with per-step human gates + 'Show me the math' receipts: the agent never spends, posts, or contacts a creator without an explicit one-click approval, and every $0.01 charge expands into the exact verified view IDs, watch-time, dedup and fraud-filter checks behind it — autonomy is something you grant in increments, not a switch you flip on day one

Key screens: Trust contract / cold open — 'Why you won't trust this agent (yet)': 7 honest objections (it'll burn my budget, post cringe, pick bad creators, hide fees, lock me in...) each with the exact control that neutralizes it; below it the plain-English 'WILL do / WON'T do / CAN'T do' table · Approval queue — agent's proposed steps as cards awaiting decision: 'Propose 12 micro-creators (~$340 est.)', 'Brief #3 caption + #ad disclosure', 'Publish Tue 6pm', 'Reallocate $80 from Creator B → Creator E'; each card has Approve / Edit / Reject + 'why the agent suggests this' · Pause & Take Over — full-bleed control: one click freezes all agent activity, shows current campaign state, lets the human run the next step manually or resume; 'kill switch' also visible as a thin persistent bar on every screen · Show me the math — a $14.27 charge expands: 1,427 verified views → list of view events, each with watch-time %, dedup status, source-anomaly score, attribution path; running total vs hard budget cap with a 'stop here' override · Off-ramp / export & leave — 'Take your campaign and go': download creators list, briefs, performance data, GMV report as CSV/PDF; cancel with no further charges; honest note on what you keep vs lose · Comparison vs the alternatives — agency / freelance VA / DIY / other 'AI tools' / SocialSeed Approval Mode across control, transparency, lock-in, hidden fees, time cost — with our own weak spots ('synthetic data in this demo', 'mock TikTok endpoints', 'no real creator escrow yet') shown in red, not hidden

Why this lens: Every other preview sells the autonomy; this one sells the brakes. It's the only surface that opens by listing why you SHOULDN'T trust the product, then earns each increment of autonomy with a concrete human-in-the-loop control — approval gates, pause/take-over, math receipts, and a frictionless exit. Trust is the product; full autonomy is the upsell.

↗ open full mockup

Persona: Solo founder / growth lead at a small TikTok Shop DTC brand who would rather *talk to* an autonomous campaign agent than configure a dashboard — comfortable delegating, wants to watch the work happen, not do it

Killer feature: An inspectable agent-at-work surface: every reply from the lead agent is backed by a live trace pane where you literally see Strategist → Sourcing → Content → Publisher → Analytics think, hand off, hit MCP tools, and ask you clarifying questions — autonomy you can audit in real time, with the hard budget cap and fraud filter visible as agent guardrails firing.

Key screens: Cold start — empty chat, single prompt: 'Tell me what you're launching and your budget.' Founder types one sentence; lead agent decomposes it into a plan card and spawns sub-agents. · Live agent trace — split view: conversation thread + right-side timeline of A2A handoffs (Strategist→Sourcing→Content→Publisher→Analytics) with each node expandable to show its Gemini reasoning and MCP tool calls streaming. · Clarifying-question moment — agent pauses, posts an inline choice card ('3 creator tiers found — approve the mid-tier shortlist, or want me to widen?') with quick-reply chips; nothing proceeds until you answer. · Brief approval card — Content agent drops an AI-generated creator brief inline (hook, storyboard, mandatory #ad disclosure highlighted) with Approve / Edit-by-chat buttons. · Verified-views ticker — Analytics agent streams results into chat: verified views counter ticking toward the budget cap, fraud-filter rejections shown as struck-through line items, GMV-attributed orders, live CPV ≈ $0.01. · Budget-cap stop — agent autonomously halts, posts a wrap-up message + ROI summary card, trace pane shows the 'hard_budget_cap' guardrail node lighting red and Publisher standing down.

Why this lens: Every other preview treats this as a product with a dashboard; this one treats the agent itself as the entire interface — the deliverable is a conversation transcript plus an auditable multi-agent trace. No nav bar, no settings page: you hire, you chat, you inspect. It leans fully into the ADK/A2A/MCP mandate by making the agent topology the literal UI.

↗ open full mockup

Persona: Operations-minded founder / growth lead at a small DTC brand on TikTok Shop who wants a campaign run for them but distrusts chatbots and wants a plain, auditable control panel — not a conversation

Killer feature: Deterministic spend ledger: every verified view is a row you can click, with the fraud-filter verdict, watch-time, source, and attribution shown — and a hard budget cap that stops the campaign at exactly deposit ÷ $0.01 views. The 'AI' is invoiced like a utility, not chatted with.

Key screens: Start a campaign — a short form: product URL, budget deposit, target market, launch window, disclosure preset (#ad locked on). Submit = hire. · Campaign status table — one row per active campaign: state (Queued / Sourcing / Live / Capped / Done), verified views, spend, remaining budget, CTR to Shop, attributed GMV. Auto-refresh on a fixed 60s interval, no animation. · Verified-view ledger — paginated table of every billed view: timestamp, creator handle, watch-time, source signal, fraud-filter verdict (Pass / Rejected + reason), cumulative spend. Exportable CSV. · Creator brief log — read-only list of the briefs the system issued (storyboard, caption, mandatory #ad disclosure line) so the brand can audit what was published in its name. · Invoice — printable A4 statement: deposit, verified views × $0.01, rejected views (not billed), budget cap line, balance returned. Plain black-on-white, print stylesheet.

Why this lens: Every other advocate will lean into the agent — chat, narration, 'watch it think'. P26 hides the agent entirely. The product is a 1998-grade SaaS ops panel: forms, tables, an invoice. Trust comes from auditability and determinism, not from a friendly voice. No emoji, no sparkles, no surprises.

↗ open full mockup