LinkedIn
X
Instagram
Threads
TikTok
YouTube
Welcome back, there 👋
Your content engine at a glance — across every connected platform.
Recent activity View all →
—
Quick start
Then sit back — it discovers ideas, writes in your voice, scores every draft, and posts (or holds for your tap).
Account checking…
—
One-click connect via the platform's secure sign-in — we use OAuth, so your password is never shared with us.
Danger zone
Switching to a different account? Reset this platform's voice, posts, drafts, and autopilot settings for a clean start. Requires typing the platform name; logged to the audit trail.
My voice checking…
—
Captured voice is injected into every draft automatically. Export the .json to reuse it elsewhere; import it to restore or move it. (LinkedIn first.)
Create a post writes · rates · rewrites · illustrates
Type a topic and hit Generate — it writes the brief, drafts in your voice, scores itself, rewrites until it clears the bar, and (for image posts) generates the image. One click; the result appears below, already gated.
The format decides the media — Single image generates one, Multi-image a few, Text / Link share none.
Video look — your default for every generated video (also in Video → clone). The director builds each reel in this style; leave on Auto to let it choose.
Advanced — write the brief yourself (optional)
Draft saved
Every draft is scored and gated automatically. Open Posts & Drafts to read the full post, see the judge's critique, regenerate, and publish.
Autopilot off
Set sources + a schedule and Autopilot runs the whole pipeline on its own: discover topics → write → rate → rewrite → (image) → auto-publish if it clears the bar, else queue for review.
Each watched post is scored for relevance to your topics; the platform acts on at most one that clears the bar every set number of minutes — reposts it, quote-tweets it, or replies with your take (your Action above) — up to your daily cap. Dedup + self-skip are automatic. Uses the free public timeline, so no special X API tier needed.
Only substantive comments (questions, real remarks) get a reply — one-word, emoji, and spam are skipped. At most one reply every set number of minutes, up to your daily cap. Dedup + self-skip are automatic. Needs an X API tier that allows the mentions timeline; it no-ops quietly if not.
…or paste a whole schedule at once
One row per post:
day, time, format. Day = Mon–Sun · time = 24h or am/pm · format = the platform's formats (the format decides the media). Header row optional.Each row = one scheduled post. Mix formats across days — e.g. an image post Monday 9am, a carousel Tuesday 10am.
Every
minutes
Runs all day, every day with no daily cap — Autopilot drips one post every set number of minutes, around the clock. The interval is the only limit, so set it to the pace you want.
Everyth post is a thread or image (alternating); the rest are single tweets.
e.g. 4 = three single tweets, then a thread or image. Set 0 for single tweets only.
Higher weight = posted more often (e.g. Single 60, Thread 30, Image 10). Set any weight and this overrides the rhythm above. Leave all blank/0 to use the rhythm instead.
Active hours: post between
and
· max
/ day (0 = no cap)
Make
% of autopilot posts a rendered video (the rest stay text/image).
Video slots stay dormant until this is above 0 and a render completes. Set the creator's clone + voice in the Video tab for a talking-head look.
Posts that clear the quality bar will publish to your live account automatically — lower-scoring drafts still queue for review. Turn on “Require Telegram approval” if you want to approve every post before it goes out.
Checking…
Scan with your phone — or tap to open Telegram — then press Start.
This connects notifications for all your platforms at once.
Bot already in the group? Paste this there instead:
This connects notifications for all your platforms at once.
Bot already in the group? Paste this there instead:
Advanced: set chat ID manually for this platform only
—
LinkedIn Engagement beta
Auto-react and comment on the accounts you watch, and reply to comments on your own posts — in your voice. This drives your real LinkedIn account via LinkedIn's private API (against their User Agreement), so it can lead to account restriction. Opt in only if you accept that risk. Publishing is unaffected — it stays on the official API.
Checking…
The safest path — no password, no cookie. Install the Social UGC extension, log into LinkedIn as normal, and engagement runs locally on your own session & IP. Pair it once with a code.
Or connect this org's session manually below.
How to find your li_at — you paste it yourself; we never see your password.
LinkedIn flags datacenter IPs and country jumps. Pick the country you actually use LinkedIn from, or bring your own residential proxy. We verify the real exit IP before saving.
Fire engagement manually to confirm it works end-to-end. These perform real actions on your LinkedIn account and need a verified session.
Or fire a single controlled write against one post (the safest first live check):
Keep engaging even when your browser is closed. One-time setup: (1) choose Use my own proxy above and save a dedicated proxy (a mobile/ISP proxy that doesn't rotate — a shared one drifts and breaks the session), then (2) open the Social UGC extension and click Enable always-on. It mints your session on that proxy so the server can run engagement round-the-clock.
Activity · all platforms 0
—
What's working · all platforms
A live rollup of your recent posts across every connected platform — posts published, average engagement where metrics are in, and your top-performing posts overall. Ranks by real engagement once a platform reports it, otherwise by the quality score. Click Repurpose on a winner (in Posts) to fan it to the others.
Top posts
—
What the AI has learned about you
Rules distilled from your own
feedback and from how your posts actually performed. A rule is scored only by the posts that
clearly beat or missed your usual engagement — an ordinary post says nothing either way — so
the counts below are the evidence so far, not a report card.
—
Video studio · brief → render
Describe a video and the production brain writes a 4-beat script, picks a style, and storyboards it — then renders a finished clip. Plan first, review the package, then render. Renders need a funded fal.ai balance.
Production packages
—
Creator clone — face & voice
—
Used for auto-generated reels when a video doesn't specify its own style/look.
—
A stored face, HeyGen avatar or cloned voice cannot be rendered without a consent record — use Record consent if you already have one from before (no retraining needed), or if you are re-consenting after a revocation. Training a face LoRA costs ~$2 on fal and takes a few minutes. Upload a dataset above, or fall back to the single reference image. A HeyGen avatar or a private ElevenLabs voice runs on this org's own key — a self-hosted LoRA
.safetensors URL does not.Paste a competitor's short-form video URL — TikTok, Instagram, X, YouTube, or Threads — and get back a structured teardown: hook strength & technique, beat-by-beat structure, on-screen text, pacing, CTA, retention risks, and what to steal for your own videos. Resubmitting the same URL reuses the cached analysis.
Recent teardowns
—
Posts & Drafts · all platforms
0 selected
—
Prompt studio · tune the agents
Override a service's base Writer / Judge / Topic instruction live — no deploy. The per-call voice, lessons, and idea are still injected on top; you're only editing the static doctrine. Saving creates a new version and makes it active across every org on that service. Revert to the shipped default anytime. Changes are audit-logged.
—
Costs · API spend across every platform
Every AI call is metered — model, tokens, and computed cost — so you can see what each model costs per platform and set the best ones. Prices are seeded estimates in each service's
pricing.py; edit there to match your actual billing. (Per-post cost lands in a later pass.)Budget & pace — used vs. your monthly cap (trailing 30d) + week-over-week
—
By platform
—
By model & role — the "which model is worth it" view
—
Top costly posts — AI spend attributed to each post (regens included)
—
Admin Console · platform-wide overview
—
Revenue
—
Recent signups
—
Key-health, budget and daily-digest alerts are delivered to the chat connected here (via the shared platform bot) — your personal DM, or a group the bot is added to. Without a connection these alerts are silently skipped.
Checking…
Scan with your phone — or tap to open Telegram — then press Start.
Bot already in the group? Paste this there instead:
Bot already in the group? Paste this there instead:
Organizations · every tenant on the platform
Change a plan, or activate / deactivate an org. Plan changes on a Stripe-active org are blocked here (they'd revert on the next webhook) — use the Stripe portal for those. Every action is audit-logged.
—
Users · account recovery
Unlock a locked account, email a password-reset link, clear MFA so the user can re-enrol, or revoke all sessions. Some actions require your own MFA. Reset / MFA-clear / force-logout are confirmed first and audit-logged.
—
Audit Log · every privileged action, newest first
—
—
Cost Watch · COGS vs MRR, this month
Per-org cost-of-goods (posts / images / videos metered) against subscription revenue for the current calendar month. Orgs are sorted by the COGS / MRR ratio — red means cost is eating the margin.
—
Health · services + credentials
Live reachability probes for the hub database and every deployed service (6 platforms + video-director + media-orchestrator), plus the health of every managed credential and proxy across services (invalid / low-credit / rate-limited). Run check → Telegram runs all mandatory checks now — key & proxy health, provider budgets, and service reachability — and DMs super-admins about anything unhealthy (the same alerts the scheduler sends automatically every ~6h).
Services
—
Credentials & keys
—
API Keys & Models · add a key once → propagates to every service that uses it
Add or replace an LLM / image / proxy / voice / messenger key — it is stored encrypted and pushed to every service that uses it, live (no redeploy). Each chip shows whether a service has the key: green = set & healthy · amber = needs attention (invalid / low credits) · grey = missing.
—
Team & plan
Your plan
Loading…
People
Invite a teammate and they get an email
with a link that signs them straight into this workspace.
Loading…
Developer console raw API responses
Responses appear here…