AfterLaunch MCP Server 是官方的吗?
是的。AfterLaunch 在 `https://afterlaunch.io/api/mcp` 运营该服务器,并发布了各客户端的配置指南。
基本信息
AfterLaunch is a remote MCP server available at https://afterlaunch.io.
https://afterlaunch.io/api/mcpclaude mcp add afterlaunch --transport http https://afterlaunch.io/api/mcpGrowth content calendar, source drafts, current move states and X production switch. Read-only. Check linkage and calendar_sync; saved words are not publication proof.
File a growth idea (optional channels, kind, score, final_text sent verbatim, desk: a post, reply or task the owner posts by hand, queued at slot_at, or a decision or read that needs the owner, due_on a date), original X draft or research outcome. Research appends a result to its idea; send expected_updated_at unchanged from the calendar. It keeps evidence holds and never authorises drafting. Requires 'act' scope; no spend or posting. Keep run_key stable. Drafts need action_kind social_post and one idea_id or source_idea; source_url is evidence. adopt_legacy only reclassifies an exact saved replay (a reply keeps its X destination).
Growth content experiments already run, newest first, with scoped evidence. Read-only; never runs one. Each is an observation, not a rate.
Confirm the AfterLaunch key works: which product run it is bound to, its scopes, tier and trial_ends_at, plus onboarding (feed_ready is false while the Growth Snapshot deep scan is still building) and credits once metering is live.
The ranked backlog of growth marketing moves prepared for this product across SEO, AI visibility (GEO), competitors and distribution: title, why, area, status, rank, whether readable full-draft material is waiting (has_draft), and draft_kind (a full draft, a brief to complete, or null). It also carries posture: who may do it, act (you can), draft_only (you draft, a person posts) or founder_voice (the founder only). Read it before improvising growth work of your own. The material BODY is not here. Live moves only unless you pass include_resolved; pass since for what is new while the deep scan adds moves. Read-only.
The automated growth marketing loops provisioned for this product (AI visibility, competitors, channel distribution) with their autonomy level, plus runnable_loop_types: what runs on demand. A runnable type only runs once a loop of that type is provisioned and enabled here.
The product run summary for this growth engine: product_url, tier, trial_ends_at, timezone, whether autonomous work is active, and delivery settings. Credits once metering is live. Read-only, free.
The drafted marketing and distribution content produced for this product, newest first: id, loop_type, kind, status and title. Read-only, free.
Read a queue draft by id. SEO linked_article {id,title,content} adds the full Markdown article.
The paste-able 'Growth (AfterLaunch)' section for this repository's CLAUDE.md or AGENTS.md: how a coding agent should use AfterLaunch for growth, marketing, SEO, AI visibility (GEO), competitor and distribution work, and which tools cost money. Free, read-only. Show it to the founder and add it with their say-so.
The measured AI visibility (GEO) results: how ChatGPT, Gemini, Perplexity and Google AI Overviews answer the tracked customer questions, plus share of voice, cited sources, competitor-owned gaps, the trend, AI crawlability and the off-site reach fold. Every reading carries its own sample count, and the sampling block says what such a count licenses, so treat a change as indicative unless it says otherwise. Read-only, free. Pass prompt_id to drill into one question.
The site-wide SEO audit: issue groups with severity and affected pages, the weekly score trend, page performance, links and authority, and day-0 search ranks. Read before any SEO work of your own; the Feed seo_fix moves are minted from it. Read-only, free. Pass include_pages for the per-page detail.
The competitor roster with the measured stat battery for you and each rival (performance, SEO, Core Web Vitals, authority, G2 and Product Hunt presence, content freshness), plus the founder's own edits. Read before any competitor or outreach work so a rival is never treated as a pitch target. Read-only, free.
What the shipped growth moves earned: tracked-link clicks per move, shipped counters, the weekly movement scoreboard, and whether GA4 and Search Console are connected. Read-only, free.
The Growth Snapshot this product run started from: positioning, the discoverability score, the actions prioritised by impact, the competitor set, and the day-0 SEO and AI visibility baselines. The stable business context, not live signal data. Read-only, free.
The product timeline: what the growth agent and the founder already did and why (drafts, ships, skips, scans, observations, decisions with their rationale), newest first. Read before acting so you never redo recent work. Read-only, free.
The Memory pages: the shared record the growth marketing engine drafts from (business basics, audience, competitors, voice samples, recent observations). Summaries with a size hint. Read-only, free.
One Memory page by slug, with its full markdown body: the record the growth marketing engine drafts from. Read-only, free.
The structured voice every AfterLaunch growth marketing draft is written in: tone, characteristic phrases, vocabulary, sentence style and what it avoids. register says which you got, and they do not sound alike: brand is the website's voice, personal is the founder's own writing, imported came from a brief. has_profile false means that voice is not extracted yet: an answer, not an error. Read-only, free.
Structured business positioning, audience, ICP, pricing, features and competitors, plus growth_strategy: the latest saved stage, bottleneck, bets and non-goals with version, source and date. Read before planning or writing. Strategy is dated judgement, not measured evidence or proof of results; check it against current founder intent. Missing facts or strategy are null. Read-only and free; never generates or revises a strategy.
The claim library: the measured numbers this founder will stand behind in public, for replies, posts and any marketing writing. Each carries the claim in one sentence, the number with its unit, keywords saying when it is relevant, and a caveat stated against the founder's own interest. READ THIS BEFORE CITING ANY NUMBER about this product, and never quote a claim without its caveat: the caveat is what makes the number safe to say. A claim marked held is NOT for use. An empty library is an answer, not an error. Read-only, free.
Put one measured number the founder will stand behind in public into the claim library, so every later reply and growth marketing post can cite it with its caveat. WHEN: a real measurement lands and the founder says they would defend it. Record the NUMBER and the caveat together; a number without its caveat is what this library exists to prevent. Never record an estimate, a projection, a rival's figure, or anything you inferred rather than measured. Free; requires the 'act' scope. Recording the same claim twice replaces your earlier wording; a pinned claim is left as it is. WHAT COMES BACK: their top moves with move ids.
Write the calibrated leans from a voice round into the FOUNDER'S OWN voice profile, the personal register, never the brand's. Two voices live here and they do not sound alike: the brand's came from the website and is what marketing pages and listings use, and this verb cannot touch it. WHEN: after a round, once an axis has enough answers to read. Send one write per round, not one per pair, and leave out an axis with too few observations: that is not a lean. Send only the axes that moved; the rest keep what they had. Free; requires the 'act' scope. WHAT COMES BACK: every axis on record, and their top moves.
Save one writing rule the person approved into their OWN voice, never the brand's. The morning marketing posts drafted for them follow it, move and voice reads carry it, and they see and remove it in Memory. WHEN: only after they approve it in words, e.g. a yes to a rule you proposed from how they edited a draft before posting. Never record one they did not approve. One sentence. At most 12. Free; requires the 'act' scope.
File one public profile or listing the person already has on a directory site, so AfterLaunch counts it as a verified presence and a weekly check confirms it is still live. WHEN: you opened the page and it is theirs. One address per site; a new address replaces the old one, the same address again changes nothing. The address must sit on that site's own host. Free; requires the 'act' scope. WHAT COMES BACK: every profile on record, the new one marked.
Add one company the person confirmed they compete with to their rival roster, which the next AI visibility scan measures them against and every competitor read carries. WHEN: only after they say yes to that rival. A rival already on the roster, by domain or by name, is left as it is. At most 10 rivals. Free; requires the 'act' scope. WHAT COMES BACK: the roster, the new rival marked.
The daily growth marketing standup in one call: what recently shipped and what AfterLaunch verified about it (receipts), the top ranked moves shippable now (ready; rank is the position among all moves, so gaps mean those ranks are not shippable now), what the engine did on its own in the last 7 days (while_you_were_gone; capped at the account's age), and whether the deep scan is still running. Relay the message field. Read-only, free.
One growth move: evidence, source material, instructions and prepared_result. linked_article is earlier work to reuse, not a completed revision. Ready means awaiting review, not published. Closed post-trial access withholds bodies without asserting absence. tracked_link_state explains a null link (no_link, unavailable, withheld); null clicks means unreadable. Read-only, free.
Accept a move only after the founder says yes. This records one durable preparation attempt and may start the configured worker only when launch_won is true. A repeat never starts another worker. Free; requires the 'act' scope.
Record completion only after the founder explicitly confirms it. Use prepared_result after reviewing the current prepared attempt, or independent_work when the founder completed it separately. Site changes need a live published_url to ship; with planned_url they stay prepared. This records an attestation; it never publishes anything. Free; requires the 'act' scope.
Apply your user's explicit decision to revise an accepted growth move's current managed attempt: its prepared or failed result is set aside and a new preparation attempt starts, only if this wins. Pass the exact job_id and expected_attempt_id. This never publishes anything. Free; requires the 'act' scope.
Apply your user's explicit decision to cancel (withdraw) unfinished work on an accepted growth move: the job is withdrawn and the move returns to pending. Pass the exact current job and attempt; legacy work also needs the exact move and legacy job states and updated_at. This never publishes anything. Free; requires the 'act' scope.
Apply your user's explicit decision to adopt work prepared before attempt tracking (a null attempt) as an accepted growth move's prepared result, ready for their review. Pass the inspected prepared_content or asset_url and the exact move and legacy job states and updated_at. This never publishes anything. Free; requires the 'act' scope.
Apply your user's explicit decision that a growth move's recorded publication is wrong: the shipped move is archived, its completion history is kept and no paid follow-through restarts. Pass the exact current job and attempt; legacy work also needs the exact move and legacy job states and updated_at. This never publishes anything. Free; requires the 'act' scope.
Dismiss a growth move (Skip in the Feed). Free; requires the 'act' scope. Optional feedback_key (closed vocabulary, see the enum) plus a free-text feedback_note explain why, which improves future marketing drafts.
Replace a pending growth move's marketing draft body before shipping. A draft carrying an unfilled placeholder or a broken link is refused rather than saved. Free; requires the 'act' scope.
Archive a growth move: a lighter dismissal than skip, with no verdict recorded. Free; requires the 'act' scope. Idempotent on an already-archived move.
Restore a just-shipped or just-skipped growth move to pending: the self-correction verb for a wrong ship or skip. Free; requires the 'act' scope. Idempotent on a move already pending. It never unwinds a real channel publish: a posted output stays posted, and an archived or expired move cannot be restored.
Resolve one drafted marketing output in the founder's review queue: action 'ship' marks it done or posted WITHOUT publishing anywhere, action 'skip' dismisses it. WHEN: call 'ship' the moment the founder confirms the content is actually out, with their go and never on your own; call 'skip' when they decide against it. Posting is the human's decision; reporting the outcome is yours, and it is not optional, because an unreported result leaves the record blind to what the work earned. Report only what is true: a draft you have written is not a draft that went out. This never posts to a channel: real publishing is a separate, human-approved egress step. Free; requires the 'act' scope and Founder tier.
Replace a drafted marketing output's body before it is shipped. Manual replacement only: the AI redraft stays in the app. A draft carrying an unfilled placeholder or a broken link is refused, and the previous draft is kept as revision history. Free; requires the 'act' scope and Founder tier.
SPENDS MONEY: run an on-demand AI visibility scan (scan='ai_visibility'), the GEO measurement of how the answer engines represent this product against its competitors. Needs the 'write' scope, a paid plan or an active trial, and an Idempotency-Key you mint. Bounded by the credit balance and the per-tenant daily and monthly caps. One scan per type per product per local day; a repeat that day replays the original result, regardless of the key.
SPENDS MONEY: run the growth-move generator on demand to top up the Feed with fresh marketing moves across SEO, AI visibility, competitors and distribution. Needs the 'write' scope, a paid plan or an active trial, and an Idempotency-Key you mint. Bounded by the credit balance and the per-tenant daily and monthly caps. Once a day per product.
Only mints a link: returns a hosted checkout_url for the Founder plan, takes no card or payment details and charges nothing itself. Payment happens solely on that hosted checkout page, which your user opens in their own browser; once they pay, the Founder-only growth controls and more metered marketing work unlock. Reachable on a free trial. Requires the 'act' scope, and it is always bound to your own enrollment.
Only mints a link: returns a hosted checkout_url for a credit top-up (pack_id small, medium or large), takes no card or payment details and charges nothing itself. Payment happens solely on that hosted checkout page, which your user opens in their own browser; the credits for metered growth marketing work land once they pay. Requires the 'act' scope and the Founder plan (not a free trial). Always bound to your own enrollment.
Which growth data sources and distribution channels are connected: GA4 and Search Console (the measurement behind SEO and outcomes), Google Business Profile, and the LinkedIn / Reddit posting channels. Each row carries an honest state: connected, not_connected, or not_available with the reason it is shut on this account. Read-only, free.
Connect a growth data source or distribution channel: mint a one-time link the FOUNDER opens in their own browser to approve it. You never complete the consent yourself; it is bound to their session. Returns url (hand it to the founder), poll_token, poll_after_seconds and expires_at; the link lasts 15 minutes. Refuses in plain words, before minting anything, when a connector is shut on this account. Free, spends nothing.
Did the founder finish connecting the growth data source you handed them a link for? Pass the poll_token you got with that link. Reports pending (wait poll_after_seconds and ask again), connected, expired (mint a fresh link) or not_found. Idempotent: nothing is spent or consumed. Read-only, free.
Disconnect a growth data source or distribution channel and delete the credential stored for it (for Google it is also revoked at Google). Idempotent: disconnecting something that was never connected changes nothing. It narrows what the growth engine can measure and where it can distribute, so confirm with the founder first. Free.
The customer questions AfterLaunch tracks: the AI visibility (GEO) question set every measurement is taken against. Returns prompts in tracking order, cap (a hard 15) and used, plus curated: false means AfterLaunch generated them and any can be replaced. Read-only, free, on every plan.
Add one customer question to the AI visibility (GEO) set AfterLaunch measures this product on. Free, spends nothing. Two honest consequences: changing the set makes the NEXT scheduled scan run at full depth rather than skipping ahead, and it re-baselines the week-over-week trend, because a comparison across two different question sets is not a real move. So add deliberately rather than churning the list. 10 to 200 characters, deduplicated, up to the included 50. Requires the 'config' scope and the Founder plan; on a free trial the questions stay readable.
Remove one customer question from the AI visibility (GEO) set, freeing a slot against the included cap of 50. Free, spends nothing. Same two honest consequences as adding one: the next scheduled scan runs at full depth, and the week-over-week trend re-baselines. Matched on the question text, ignoring case and punctuation, and IDEMPOTENT. Requires the 'config' scope and the Founder plan.
Did the growth work move anything? The proof read for a period: AI visibility (GEO) at the START and the END of the window, the moves shipped between them, the scans that ran, and the receipts with what was verified live and the clicks each earned. ALWAYS read like_for_like before reporting a change: true means both scans asked the same tracked questions so the movement is real, false means the questions or answer source changed in the window, so it is a new baseline rather than a rise or a fall, and null means it cannot be proven either way. Never present a delta as a result when like_for_like is not true. Read-only, free.
Set aside one AI visibility (GEO) gap the founder is not going to act on: a tracked question where rivals get named and this product does not. The same verdict as a skip, on the AI visibility surface, and it resolves the gap in ONE place: it stops being re-proposed here AND on the growth Feed, and the reason improves what gets ranked next. Free; requires the 'act' scope, and available on every plan because telling the agent "not this one" is never a paid feature. Idempotent: a gap already shipped or archived is left as it is.
Write one durable thing you have learned about this founder or their product into AfterLaunch's Memory, the record every growth and marketing draft is written from. WHEN: the moment the founder tells you something that will still be true next month and that AfterLaunch could not have measured itself. A standing preference, a rule about how they write, a fact not on the site yet, or what a shipped move actually earned. Recording it is part of the job, not a favour: AfterLaunch drafts from Memory, so anything you keep to yourself is a correction the next draft will need again. Record what LASTS. Never record chatter, a passing mood, a restatement of the Snapshot or Memory, or anything you inferred rather than heard. One sentence in the founder's own terms, 10 to 500 characters. Free; requires the 'act' scope. Recording the same insight twice replaces your earlier note; a pinned note is left as it is. WHAT COMES BACK: their top moves with move ids, so recording hands you your next instruction without a second call.
Add a pending move to your user's growth moves. WHEN: you found a thread on a community they work, or their plan calls for a placement ask, outreach, a directory listing or a page for their site, and no move covers it yet. A repeat (same kind and target, or kind and title) hands back the existing move id. It comes back with its posture: a Reddit or Hacker News target is draft_only, meaning you draft and your user posts, and you never post there yourself without their yes in your thread. Outreach goes from their address, never without their yes. Free; 'act' scope; a fixed number a day. WHAT COMES BACK: the move id and posture.
Unclaimed preparation jobs, oldest first. Each row carries job_id and its nullable attempt_id. Queue summaries never include prepared_content. Free; requires the 'act' scope.
Try to claim one queued preparation attempt. Pass job_id and the exact nullable attempt_id of a queued job. Start work only when the response says claim_won:true. An unchanged result is a non-winner and must not start work. Free; requires the 'act' scope.
Report a claimed attempt with its exact attempt_id: running; needs_input with one question for your user; shipped with the live published_url, or planned_url once merged (we check it); ready_for_review with prepared_content, or asset_url when your user must post it or when it is the pull request you opened; failed with fail_reason. Ready_for_review can be reported again, improved, until your user uses it. If you cannot read the whole thread, say so and ask your user for a screenshot. Free; requires the 'act' scope.
AfterLaunch MCP Server 把 Claude、Codex、Cursor 或 VS Code 连接到 AfterLaunch。AfterLaunch 是一个面向 AI 搜索和 SEO 的增长引擎,衡量 ChatGPT、Gemini、Perplexity 和 Google AI Overviews 如何提及你的产品,并对能改变这一结果的增长动作进行排序。
AfterLaunch 会扫描你的网站和竞争对手,然后生成一个由小型营销动作组成的排序看板,例如认领目录收录、回复社区帖子,或者针对竞品被提及而你没有被提及的问题发布一个页面。通过 MCP 服务器,你自己的编程智能体可以读取这些动作、起草工作内容,并在你确认后将其记录为已发布。该服务器提供 55 个工具。
你需要一个已扫描产品的 AfterLaunch 账号:在 https://afterlaunch.io 运行免费扫描(无需绑卡,完整扫描约需十分钟)。编程智能体使用 API 密钥认证,在 https://afterlaunch.io/app/connect#keys 创建,并以 Bearer 令牌发送。聊天类应用则通过 OAuth 登录,无需创建密钥。
端点:
https://afterlaunch.io/api/mcp
先在 shell 中导出 AFTERLAUNCH_API_KEY,然后运行:
claude mcp add afterlaunch --transport http https://afterlaunch.io/api/mcp --header "Authorization: Bearer $AFTERLAUNCH_API_KEY"
打开 Customize、Connectors,添加一个 URL 为 https://afterlaunch.io/api/mcp 的自定义连接器。然后在 Connectors 中打开 AfterLaunch,点击 Connect 并在浏览器中批准。在 Team 或 Enterprise 套餐中,由 owner 在组织设置中统一添加一次。免费 Claude 套餐允许添加一个自定义连接器。
把以下内容添加到 .cursor/mcp.json(或 ~/.cursor/mcp.json 以应用于所有项目),并在打开项目前设置 AFTERLAUNCH_API_KEY:
{
"mcpServers": {
"afterlaunch": {
"url": "https://afterlaunch.io/api/mcp",
"headers": {
"Authorization": "Bearer ${env:AFTERLAUNCH_API_KEY}"
}
}
}
}
把以下内容添加到 .vscode/mcp.json。VS Code 会在首次启动时提示输入密钥,并保存在密钥存储中:
{
"servers": {
"afterlaunch": {
"type": "http",
"url": "https://afterlaunch.io/api/mcp",
"headers": {
"Authorization": "Bearer ${input:afterlaunch_key}"
}
}
},
"inputs": [
{
"type": "promptString",
"id": "afterlaunch_key",
"description": "AfterLaunch API key",
"password": true
}
]
}
codex mcp add afterlaunch --url https://afterlaunch.io/api/mcp --bearer-token-env-var AFTERLAUNCH_API_KEY
AfterLaunch 还在 https://afterlaunch.io/agents/onboard.md 中提供了 Windsurf、Gemini CLI 和 Cline 的配置。Cline 需要设置 "type": "streamableHttp",因为服务器不提供旧版 SSE 传输。
get_standup 一次调用给出近期发布内容和排名靠前的动作;list_feed、get_move、accept_move、ship_move、skip_move、review_move、undo_move 和 propose_move 用于管理看板。get_visibility、get_seo、get_competitors、get_outcomes、get_scoreboard 和 get_snapshot;list_tracked_prompts、track_prompt 和 untrack_prompt 管理被测量的客户问题。get_business_context、get_voice_profile、get_claims、record_claim、record_insight、list_kb_pages 和 get_kb_page。list_outputs、get_output、redraft_output、set_output_status、get_content_calendar 和 record_content。whoami、list_connections、start_connection、poll_connection 和 get_setup_snippet(生成可放入 CLAUDE.md 或 AGENTS.md 的段落)。refresh_scan 和 replenish_feed 会产生费用;create_checkout 和 create_topup 返回供创始人打开的付款链接。whoami 会报告 feed_ready 为 false。"type": "streamableHttp";缺少 type 时会被当作 SSE,而服务器不提供 SSE。AFTERLAUNCH_API_KEY。refresh_scan 和 replenish_feed 需要 write 权限范围,以及付费套餐或有效试用。是的。AfterLaunch 在 `https://afterlaunch.io/api/mcp` 运营该服务器,并发布了各客户端的配置指南。
扫描免费,并开启 14 天无需绑卡的试用。Founder 套餐为每月 $149。
Claude Code、Codex、Cursor 和 VS Code 等编程智能体使用 API 密钥作为 Bearer 令牌。claude.ai 等聊天应用通过 OAuth 登录,不需要密钥。
AfterLaunch 的工具在发布动作前需要创始人确认,会产生费用的工具也有明确标注。
免费扫描覆盖 ChatGPT 和 Perplexity;Founder 套餐增加 Gemini 和 Google AI Overviews。
AfterLaunch 网站将 ChatGPT 列为支持的客户端,以聊天应用的方式连接。
AfterLaunch 的远程 MCP 服务器是一个托管的 Model Context Protocol 端点,地址是 https://afterlaunch.io/api/mcp,AI 助手无需在本地安装或运行任何东西即可连接。
把端点 https://afterlaunch.io/api/mcp 添加到任意支持 MCP 的客户端,比如 Claude Code、Cursor 或 VS Code。本页的接入代码块可以一步配置好每个客户端。
此收录尚未确认 AfterLaunch 的认证要求。连接前请查阅提供方文档。
AfterLaunch 暴露的是 Streamable HTTP 端点,这是远程 MCP 服务器普遍使用、主流客户端都支持的传输方式。
One MCP connection for all your Gmail, Google Calendar, Drive and Contacts accounts, Slack workspaces and Notion — free.
App Store keyword research, rank tracking, competitor analysis and authorized performance data.
Search and read public data from X/Twitter, Reddit, YouTube, Instagram, TikTok, Facebook, Threads, Bluesky, Trustpilot, Amazon, Google Maps, news, forums and the web. 90+ read-only tools plus ready-made research playbook
AOI is a commercial environmental-intelligence gateway that gives AI agents programmatic access to 38 real-world intelligence capabilities — climate risk, satellite monitoring, methane detection, water & soil, vegetation
AI Video MCP by AITuber is a hosted MCP server that lets AI agents create and publish video. You connect it once, then ask your agent for a video in plain language. It writes the script, picks a voice, generates the visu
评论