Your Threads workflow, inside your AI
Your assistant reviews the account, prepares the next posts and replies, then follows the publishing mode you choose: review first or fully automated within your rules.
Start with approval required, then grant direct publishing when you are ready. Timed scheduling, automatic voice learning, and a hosted built-in writer remain roadmap items.
Interactive permission preview
Your assistant prepared
Most people quit a posting habit in week three. What kept me going was writing five drafts on Sunday and letting the week spend them.
Choose your automation level
Change any switch. Start read-only, keep review on, or turn review off for full automation within the rules you set.
Your Threads loginNever shared
Your permissions and publishing mode decide what crosses
The job
Your assistant can use the account context this product makes available instead of a prompt assembled from screenshots and copied posts. It inspects your own recent content, replies, and available insight snapshots before preparing the next action.
Read your profile, recent posts, replies, conversations, and the insight fields Meta returns.
Draft in the assistant you already use, then propose a post or reply through the same connection.
Give the next session account-relative medians, sample sizes, and labelled experiments without inventing a universal score.
Choose your assistant
The repository has a recorded read-only interoperability check. Start with the guided local setup.
A synthetic-account check discovered and called the gateway. Consumer Claude clients are not implied.
The desktop app has an official MCP setup path. ChatGPT web needs a published plugin that is not shipped yet.
The loop available now
Your assistant reads the account data and available snapshots the official API returns.
The writing happens in the assistant you chose. Nothing forces a generic built-in writer between you and that assistant.
Anything requiring approval waits in one review queue and is checked against the current rules again before publication.
The assistant can compare a labelled group with the same account’s baseline, with every sample size and limitation attached.
Control, when you want the detail
Begin read-only. Add post or reply capability later. Keep approval required or raise autonomy within rules you set. The assistant never receives the Meta credential, and disconnecting it does not require breaking the Threads account connection.
Read the security and credential details or inspect the MCP and API surface.
What is live, and what comes next
The current workspace includes bounded niche discovery, source-preserving draft requests, reviewable creator and voice context, assistant setup, a timezone-aware calendar, focused inbound replies, and evidence-based workflow progress. Automatic timed publishing, inferred voice learning, and second-cycle learning remain roadmap items.