The X (Twitter) MCP server for AI agents.

Drafts, AI composing in your voice, repurposing videos and posts, scheduling, queue slots and instant publishing.

repurpose_source

Repurpose studio: turn a YouTube video (its captions), a podcast episode page with a transcript or show notes, a blog post, a newsletter issue or pasted text into a 5 to 8 part thread and five standalone posts with no links, plus an X Article that credits the source when withArticle is true, all in the account's voice. Checks every draft for lines copied from the source (and rewrites them), near-copies of known viral posts, misquotes and figures the source doesn't give. Nothing is posted. Costs credits; the Article costs extra.

list_repurpose_jobs

Recent Repurpose studio runs for the active account, newest first: the source, its kind, how many drafts it gave and how many originality warnings are left.

get_repurpose_job

One Repurpose studio run: the source, the thread, the five posts, the X Article and the originality check.

list_scheduled_posts

List pending scheduled posts.

schedule_post

Schedule a draft for future publication.

cancel_scheduled_post

Cancel a pending scheduled post.

publish_now

Publish a draft immediately.

get_queue_slots

List the manual posting-queue time slots.

set_queue_slots

Replace the manual posting-queue time slots.

list_best_times

List data-driven best posting times.

list_drafts

List drafts.

create_draft

Create a draft (post / thread / article).

get_draft

Get a draft.

update_draft

Update a draft.

delete_draft

Delete a draft.

share_draft

Share a draft with a client (Scale): a private link, valid 30 days unless expiresInDays says otherwise, where they read the draft, comment and approve it or ask for changes without an Xpert login. The link is returned once; only a hash of it is kept, so it can't be shown again.

revoke_draft_share

Turn off one of a draft's share links. It stops working at once.

get_draft_collab

A draft's client review (Scale): its share links (never the link tokens), the approval state and who decided, comments from the team and from clients, and the activity log of who changed what, with AI assistant (MCP) and API edits kept apart from people's.

comment_on_draft

Add a team comment to a draft (Scale). Pass shareId to show it on that share link so the client sees it too; otherwise only the team does.

resolve_draft_comment

Mark a comment on a draft as resolved (Scale).

autopilot_draft

Agentic draft loop: generates a post/thread in your voice, predicts its hit score against YOUR history, applies its own critique (stronger opener + weak-span rewrites), re-scores, and does one targeted AI revision if still under targetScore. Returns the final draft, before/after scores, the prediction, and the step-by-step trace.

compose_post

Generate post text with AI in your voice. Returns the generated text.

rewrite_text

Rewrite text with AI. Returns the rewritten text.

predict_post

Predict how a draft post will perform: an engagement estimate from your own most-similar past posts plus an AI hook critique.

viral_check

Will this post go viral? One verdict covering every pre-publish case: format rules (links, hashtags, bait, length, duplicates), your own history, what already went viral in your niche, timing and account size, with a 0-100 score, a probability, and a fix per negative factor.

schedule_spacing

Smart spacing (Growth): the account's other posts within a feed session of a time, the share of its score a bunched post keeps under X's author diversity decay, the next clear time, and the new-author window for accounts under 1,000 followers.

voice_match

How much a draft sounds like the account's own posts: a 0-100 voice match against its closest posts, how AI it reads, and the lines that drift from its voice with why. Free.

preflight_check

The checklist before posting: voice match, AI tells, mute and report risk, creator payout safety (engagement asks, copied posts), originality, the link price and the AI-image label, each pass, warn or fail with a fix. Free.

fact_check

Check a draft's factual claims (numbers, dates, names, events, quotes) against the web: supported, disputed or unclear, with a note and sources per claim. Costs credits.

hook_lab

Hook Lab: twelve opening lines for one idea across six angles (story, contrarian, specific, question, list, lesson) in the account's voice, ranked by how likely each is to stop the scroll and how much it sounds like the account, with why each could work. Costs credits.

hook_lab_pick

Record which Hook Lab opening the user chose, next to the one ranked first, so Edit memory learns their taste. Free.

get_model_match

Best-model match for the active account: which AI model writes most like the user, from a monthly test where each candidate rewrites up to ten of their real posts from the core idea and Jev picks the closest. Returns each model's wins, trials and mean closeness (0 to 1), the winner and whether drafts use it, when the test ran, the model drafts use now and the council lineup. enabled is false on plans without it (Growth and up).

run_model_match

Run the best-model match now instead of waiting for the monthly run (Growth, once a day per account). Takes about a minute; the model usage is charged at cost, about 100 credits. Returns the new ranking.

council_draft

Council drafting (Scale): three or four top models each draft the brief as a post or thread in the account's voice, and Jev ranks the drafts on voice match, reach (would it stop the scroll) and how human it reads. Returns every draft best first with the three reads (0 to 1), the best two with a one-line reason. Drafts only, publishes nothing. Costs credits.

get_voice_tune

Auto-tuned voice (Scale): the voice prompt setup kept this week (which example posts and learned rules it carries), its closeness score against the default setup's (0 to 100), and every setup tried.

strategy_coach

Get an AI growth action plan for the next 7 days, grounded in your real analytics.

list_suggestions

List pending AI post suggestions.

refresh_suggestions

Generate missing AI post suggestions now.

post_suggestion

Publish a suggested post immediately.

dismiss_suggestion

Dismiss a suggested post.

list_templates

List the org's reusable tweet templates, most recently edited first.

create_template

Create a reusable tweet template (used in compose and as the auto-plug CTA source).

get_template

Get one tweet template.

update_template

Update a tweet template (partial).

delete_template

Delete a tweet template (also detaches it from any auto-plug).

Your voice guide and Voice DNA, the take bank and interviews, voice notes, and named voices for a brand or a client.

list_voices

The active account's named voices (Scale): up to five, such as founder, brand and Articles, each with where it is used by default (any, post, thread, reply, article), its instructions, rules and example posts, and which one is the default. Pass a voice's id as voiceId to compose, autopilot, hook_lab, council_draft or create_draft to write in it.

create_voice

Add a named voice to the active account (Scale, five at most): a name, where it is used by default, how it differs from the account's own voice, hard rules and example posts. Drafts in it keep the account's learned voice and follow these on top.

update_voice

Change a named voice: its name, where it is used, instructions, rules, example posts or whether it is the default.

delete_voice

Delete a named voice. Drafts that used it go back to the account's own voice.

set_default_voice

Make a named voice the account's default: drafts use it whenever no other voice is picked or set for the kind of post.

get_style_profile

Get the writing-style profile for the active account.

update_style_profile

Save manual edits to the writing-style profile.

reanalyze_style

Re-derive the style profile from cached posts.

regenerate_style_brief

Regenerate the account's About brief from live signals (bio, site, likes, bookmarks, lists).

get_voice_guide

How to write as this account, as one markdown document: the account brief, voice rules and Voice DNA, rules learned from the user's edits, their own takes, their best recent posts and the writing rules. Read it before drafting anything in the user's voice, then run preflight_check on the draft.

get_voice_dna

The account's Voice DNA: measured writing habits per context (posts, threads, replies: length, rhythm, line breaks, casing, punctuation, emoji, openers, endings, signature words, words never used), the humour and stance read, and the user's corrections.

correct_voice_dna

Switch a Voice DNA trait off or on, replace a word list (signature words, words never used, openers) or reset it to what was measured. key is '<context>.<trait>'.

search_my_posts

Semantic search over your own posts.

list_learned_rules

Edit memory: the writing rules Xpert learned from how you edit its drafts, strongest first, each with a before and after example and whether it shapes drafts now, plus how many draft edits it has recorded.

remove_learned_rule

Remove a rule Xpert learned from your edits. It stops shaping drafts and is never learned again.

voice_interview_status

The voice interview for the active account: the interview in progress (its transcript and the question waiting for an answer), whether the account has done its interview, whether this week's check-in is due, how many takes it has, and what an interview and a check-in cost.

start_voice_interview

Start the voice interview (kind onboarding: about nine questions on the user's work, stories, opinions, numbers, lessons and words) or this week's three-question check-in (kind checkin), or resume the one in progress. Returns the first question. The first completed interview is free; later ones and each check-in cost credits. Put the questions to the user and send their own answers; never answer for them.

answer_voice_interview

Send the user's answer to the current interview question (typed, or the transcript of a spoken answer). Returns the next question, or when the interview is done, the takes pulled from it. Only send what the user actually said.

finish_voice_interview

End the interview early and pull takes out of the answers so far. An interview closed before any answer costs nothing.

list_takes

The take bank: the user's own stories, opinions, numbers, lessons and beliefs, newest first, with counts per kind and how often drafts used each. Drafts draw on the closest takes instead of filler.

add_take

Add a take the user stated in their own words: a story, opinion, number, lesson or belief (kind is guessed when left out). Never invent one.

update_take

Edit a take's text, kind or topic.

archive_take

Archive a take: drafts and the pre-flight stop using it.

voice_note_to_posts

Turn a voice note into a post, a thread of 3 to 6 parts and two reply-ready takes, in the account's voice. Send what the user said as transcript (their words, never invented). rewrite runs from 0 (keep their words, only filler cut) to 100 (full rewrite for the feed); facts, numbers and opinions are kept at every level. Apps can send the recording instead, as multipart audio. Costs credits.

list_voice_notes

The active account's recent voice notes, newest first: each transcript, its rewrite level and the post, thread and takes drafted from it.

redraft_voice_note

Draft a voice note again at a new rewrite level (0 keeps the spoken words, 100 is a full rewrite). Returns the note with its new post, thread and takes. Costs credits, less than a new note.

Account metrics, growth history, content genome, the weekly what-worked read, engagement prediction and calibration.

analytics_overview

Account snapshot: followers, posts and 24h metric deltas.

analytics_growth

Follower/growth chart data.

analytics_activity

Daily activity buckets (posts, views, likes, engagement rate).

analytics_content

Content effectiveness stats and top posts.

post_metrics_history

Metric history for a single post.

analytics_realtime

Real-time metrics for the latest own posts.

follower_history

Daily follower-count history (gained/lost/net).

analytics_calibration

How well the AI predictor has matched reality for this account: measured samples, mean bias of actual vs predicted engagement rate (pct points), and share of predictions within 1pp.

analytics_genome

Content genome: your recent posts clustered into up to 6 labelled themes with post count and average engagement rate per theme, best first. Cached 24h.

weekly_read

The weekly what-worked read for the active account: last week's posts against the account's own previous weeks, what worked, what didn't, one thing to try next, underdog posts worth rewriting and how close the hit scores came to reality. On Growth it adds whoResponded: which follower segments reposted, replied to or quoted last week's posts against their share of the followers, with a one-line lesson (null until there is enough to say). Also lists earlier weeks. Written each Monday for a week with posts.

get_weekly_read

One week's what-worked read, by the Monday (UTC) the week began, as YYYY-MM-DD.

mark_weekly_read_seen

Mark a weekly read as seen, which also clears its in-app notification.

list_post_autopsies

Per-post autopsies for the active account, newest first: each original post a day or two after it went out, against the account's own usual (views, engagement rate and weighted engagement as multiples of usual, percentiles, the pre-post hit score against where it landed), a verdict (outperformed, typical or underperformed), Jev's hook, format and topic tags and a two-line lesson. On Growth, segments says which follower segments engaged against their share of the followers, with a one-line lesson (null until there is enough to say). Pass ids to get specific posts.

get_post_autopsy

One post's autopsy, by its X post id: its numbers against the account's usual, the verdict, tags, the two-line lesson and, on Growth, which follower segments engaged.

winning_formats

The account's own winning formats from the last 60 days of post autopsies: hook, format and topic combinations and media or length groups whose posts beat its usual engagement (median, at least 3 posts each), best first, with example posts. Build new posts on these rather than on everyone else's.

list_posts

List your own posts with metrics.

lookup_post

Look up ANY public X post by its id: text, author, media and metrics. Powers embed previews and post research.

list_post_repliers

Everyone who replied to a post, as an enriched outreach dataset: handle, bio, followers, their reply and a link, most-followed first, plus a downloadable CSV URL for export.

Read your home timeline, see what your network is amplifying, find the highest-leverage posts to reply to, and work the For you feed.

get_timeline

Read the home timeline (you + everyone you follow), newest first. Paginate older posts by passing `cursor` (an ISO date from the previous page's next_cursor).

draft_timeline_reply

Generate a voice-matched reply, remix, or quote for a post. Returns the drafted text only: it does not publish.

post_timeline_reply

Publish a reply, quote, or standalone remix straight to X. For reply/quote, pass the target `tweetId`.

list_foryou

The For You feed: pre-drafted, voice-matched replies to fresh rising posts from your network. Each card carries the target post and a ready reply draft.

dismiss_foryou

Dismiss a For You reply card.

regenerate_foryou

Regenerate the drafted reply on a For You card (metered), optionally built on your take and in one reply style.

post_foryou

Publish a For You reply to X now, optionally with edited text. Only for cards with replyVia 'api' (the post mentions the user); X refuses other replies from apps, so this answers 409 reply_needs_you with an intentUrl: use handoff_foryou.

schedule_foryou

Schedule a For You reply for later: pass when = 'next_slot', 'best_time', or an ISO datetime. Only for cards with replyVia 'api'; others answer 409 reply_needs_you.

handoff_foryou

The user sends a For You reply from their own X (X only takes most replies that way): returns intentUrl, X's composer pre-filled with the reply, records the text and takes the card off the feed. Give the user the link; nothing is posted by Xpert.

network_pulse

What's resonating across the accounts you follow right now: top posts of the last 48h ranked by engagement rate, plus the current author leaderboard.

network_authors

Leaderboard of the followed authors driving engagement in your feed, with each author's current top post.

network_opportunities

Reply opportunities: fresh posts from your network that are semantically in your lane AND still gaining engagement, the highest-leverage posts to reply to right now.

network_brief

AI intelligence brief on your corner of X: themes, momentum, and a concrete playbook grounded in what your network posted. Cached ~12h; pass refresh=1 to regenerate (metered).

sync_network

Re-embed new timeline posts and refresh the network author rollup now (normally hourly).

audience_interests

Semantic search over your FOLLOWERS' bios: who in your audience cares about a topic, with their bio and reach.

Keyword monitors, intent-scored lead signals, reply suggestions, the approval queue, CRM contacts and alerts.

get_reply_budget

The daily reply budget for the active account: the replies a day the user set (null when off) and how many replies they handed to X or sent today (UTC). A pace the user chose, never a block.

check_reply

The free check before a reply goes out: whether it reads like AI (Jev's slop read and phrase tells) and, with postId, whether it closely repeats a reply already under that post. Returns one flag and its line; it never blocks a send.

shield_posts

AI Shield: for up to 20 posts (replies, mentions), how likely each was written by AI and how likely it came from an automated account, 0 to 1, with a label: 'automated' at 0.7 on the bot question, else 'ai' at the slop detector's verdict bar, else null. Probabilities, not proof. Free, capped per user.

list_signals

Leads found by your keyword monitors: posts the intent classifier scored as commercial intent (shopping / frustrated / curious), best first. Reply via the engagement item endpoints.

list_contacts

The org's CRM contact list (leads, engagers, partners) with tags and notes.

get_contact_by_handle

Look up a CRM contact by X handle.

add_contact

Add or update a contact by handle, with optional tags, notes and source.

add_contacts_batch

Save up to 50 X users to the org's Contacts CRM in one call (upsert by handle; existing notes are kept unless new ones are given). The assistant's Fresh-leads card uses this.

update_contact

Update a contact's name, tags or notes.

delete_contact

Delete a contact.

list_monitors

List engagement monitors.

create_monitor

Create an engagement monitor (mentions / keyword / competitor).

delete_monitor

Delete an engagement monitor.

list_engagement_items

List engagement items across monitors. sort=priority returns the reply-triage order: each mention carries triageLabel (question|praise|criticism|lead|spam|troll|other) and triagePriority 0-100 (how much it needs an answer).

suggest_engagement_reply

AI-generate a reply suggestion for an engagement item, optionally built on your take and in one reply style.

respond_to_engagement

Queue a reply to an engagement item for approval. Only for items with replyVia 'api' (a mention); others answer 409 reply_needs_you with an intentUrl: use handoff_engagement_item.

handoff_engagement_item

The user sends a reply to an engagement item from their own X: returns intentUrl (X's composer pre-filled), records the text and marks the item answered. Nothing is posted by Xpert.

ignore_engagement_item

Mark an engagement item as ignored.

list_notifications

In-app alerts raised by the background jobs: a post in its first-hour amplify window, an autopilot draft held for review, a paused campaign.

read_notification

Mark one notification as read.

read_all_notifications

Mark every notification as read.

list_dm_campaigns

List DM outreach campaigns.

create_dm_campaign

Create a DM outreach campaign (bulk messaging).

get_dm_campaign

Get a DM campaign and its targets.

launch_dm_campaign

Queue a DM campaign's targets for approval.

pause_dm_campaign

Pause a DM campaign.

get_automation_settings

Get automation settings for the active account.

update_automation_settings

Update automation settings (CTA, auto-DM, auto-delete).

set_digest_prefs

Set daily-digest (morning prep) preferences: local hour and timezone (null hour opts out).

list_approvals

List approval-queue items.

approve_item

Approve a queued item. A reply X won't take from apps answers 409 reply_needs_you with an intentUrl: use handoff_approval.

handoff_approval

The user sends a queued reply from their own X: returns intentUrl (X's composer pre-filled) and marks it sent. Nothing is posted by Xpert.

reject_item

Reject a queued item.

bulk_approvals

Bulk approve or reject queued items.

Semantic search over public posts, the viral-post library, competitor autopsies and profile x-rays.

list_competitors

List the competitor handles tracked for the active account (max 3).

add_competitor

Track a competitor handle (hard cap 3 per account). The weekly autopsy analyses their top posts against your own positioning.

delete_competitor

Stop tracking a competitor.

competitor_analysis

The cached weekly competitor autopsy: per-competitor observations of what's working for them, each with a concrete angle for you.

get_profile_xray

The cached Profile X-ray of any X account: voice, topics, what works, hook formulas, cadence, weaknesses, and how to get their attention. 404 until analyze_profile has run.

analyze_profile

Run (or re-run) a Profile X-ray on any public X account: pulls their recent posts, computes engagement stats, and produces a deep AI analysis cached for the org. Metered (credits).

browse_viral_library

Browse the global viral-post library; filter by niche/hook/format, sort by score or recency.

list_library_niches

List the niches the viral library covers.

search_viral_library

Semantic search across the viral library: find high-performing posts similar to a topic or to your own post.

discover_posts

Semantic search over a shared, growing corpus of public X posts: find high-performing tweets on any topic to riff on. Each search also backfills fresh results from the firehose into the shared corpus.

search_users

Find X ACCOUNTS by keyword (bio/name match): competitor discovery, prospecting, contact enrichment. Returns profiles with follower counts.

find_people

Find real X accounts matching an audience description: authors of posts matching your queries (active on the topic right now) merged with bio matches, deduped and enriched with bio, followers, and a sample post. The caller judges relevance and picks the keepers.

find_leads

Autopilot customer finding: checks your signal agents (keyword monitors) for fresh high-intent catches first, falls back to a live X search, and AI-qualifies every candidate against what you sell. Returns only fits (≥60) with a one-line reason each, plus whether each is already a contact.

x_list_posts

Tweets from a curated X List by list id: plug a List in as an inspiration source.

community_posts

Tweets from an X Community by community id: niche-community inspiration source.

Follower analytics, bot detection, relationship history with any handle.

audience_segments

Audience segments (founder / developer / marketer / creator / investor / …) over follower bios with counts and average churn risk, plus how many followers are at high risk of unfollowing.

mutuals_radar

Mutuals radar (Growth): the mutuals most likely to reply, ranked by how often they engage with the account's posts. A mutual's reply on an original weighs 20 instead of 5 in X's ranking.

audience_overview

Audience overview: follower counts (active, new this week, likely bots, suspicious) and recent followers.

sync_followers

Start a background sync of your followers, with bot scoring.

list_bot_followers

List flagged likely-bot and suspicious followers (worst first).

scan_bot_followers

Run an AI pass to re-judge borderline 'suspicious' followers using name/bio context.

mark_follower_removed

Mark a follower as removed (after you block or remove them on X).

block_follower

Block a follower on X via the API, then mark them removed.

unblock_follower

Unblock a follower on X (reverses block) and clear their removed flag.

get_relationship

Relationship/CRM status for a handle: whether they follow you, since when, bot risk, and how many of your posts they've engaged.

Long-form X Articles from blocks, AI cover images, image generation, GIFs and media uploads.

text_to_image

Generate an image from a text prompt (default flux schnell). Pass `model` to pick one of your plan's catalogue models (GET /v1/models; each has its own credit price), plus `aspect`, `negativePrompt`, `steps`, `guidance` or `seed`. Returns a public image URL.

image_to_image

Restyle or vary a source image (`imageUrl`) from a prompt. `strength` (0..1) controls how far from the source to travel. Returns a public image URL.

search_gifs

Search Tenor for GIFs to attach to a post. Returns preview + attachable URLs.

attach_gif

Stage a picked GIF: fetches it server-side, stores it on the Xpert CDN, and returns a url to put in a draft block's mediaUrls (uploaded to X at publish).

generate_article_image

Generate an AI image with Recraft V3 and upload it to X: kind 'banner' renders the title as bold poster typography (for covers), 'inline' makes a no-text section illustration. Returns the X mediaId to attach plus a preview URL.

upload_article_image

Upload an image for the article editor (multipart 'file'); returns a CDN URL to embed.

resolve_article_media

Resolve an Xpert CDN image URL to an X mediaId for attaching to an article.

import_article

Pull a PUBLISHED X Article (anyone's) by its tweet id, mapped to editor blocks: competitor/article research.

create_article

Create a long-form X Article from blocks (text / heading / image-with-mediaId), with optional AI cover via coverMediaId. Set publish=true to go live immediately; otherwise it stays a draft on X.

upload_media

Upload an image to X (provide an image URL or base64). Returns a mediaId.

X account management, the product catalogue for plugs, API keys, onboarding and Learning mode.

get_me

Get the current user profile.

get_x_account

Get the active X account (auto-connects if needed).

list_x_accounts

List all X accounts you can act as.

set_autosync

Enable or disable auto-sync for the active X account.

resync_account

Reset the active X account and re-ingest its data.

disconnect_x_account

Disconnect an X account and delete its cached data.

list_api_keys

List active API keys.

create_api_key

Create an API key (the secret is returned once).

revoke_api_key

Revoke an API key.

onboarding_status

Check onboarding / ingest progress.

start_onboarding

Kick off background profile sync (requires a connected X account).

get_x_connect_url

Get a URL to connect a new X account (open it in a browser to finish).

learning_profile

Learning mode (opt-in, Settings → Learning): what Xpert has learned about how this user works and writes, summary, likes, dislikes, working style, suggested prompts, compose defaults.

set_learning_mode

Turn Learning mode on or off for the current user. Off by default; when on, the web app records clicks, navigation and writing to personalise the assistant and composer.

rebuild_learning_profile

Rebuild the Learning-mode profile from the last 30 days of activity right now (normally nightly). Needs Learning on and some activity.

clear_learning_data

Delete everything Learning mode has recorded and derived for the current user. Leaves the on/off setting unchanged.

list_products

List your products.

create_product

Add a product (name + optional URL). The URL is scraped to auto-fill its description.

get_product

Get one product.

update_product

Update a product.

rescrape_product

Re-scrape the product's URL to refresh its description.

draft_product_post

Draft an X post about the product, in your voice. Returns the generated text.

delete_product

Delete a product.