TikTok MCP server: post from your AI agent, with your approval

A TikTok MCP server lets an AI agent such as Claude, ChatGPT or Codex work with TikTok through tools. Flaghand is one for posting: your agent drafts TikTok photo and video posts, checks them and asks you. You approve each exact post with your passkey, and Flaghand posts it then, or at its time. TikTok hasn't audited Flaghand's app yet. Until it does, a direct post works only for a private TikTok account, as Only me. The other way is sending the post to your TikTok inbox as a draft, which you finish and publish in the TikTok app.

It's not for reading TikTok: other people's videos, search, trends, analytics or scraping. It doesn't run TikTok ads either. For those jobs, see other TikTok MCP servers below.

Until TikTok audits the app

TikTok limits apps it hasn't audited. Flaghand's TikTok app is unaudited, so today:

• Posting directly works only for a private TikTok account, with who can view set to Only me (SELF_ONLY). Choosing Only me on a public account isn't enough.
• Sending to the inbox works for every account. TikTok puts the post in your TikTok inbox as a draft. You finish it and publish it in the TikTok app, where you choose who can view it.

When a direct post can't go out, Flaghand's check offers the inbox instead. Flaghand doesn't promise a public TikTok post until TikTok audits the app.

Set it up

The server is https://app.flaghand.com/mcp. Add it to your app:

• Claude (Free, Pro and Max; Free allows one custom connector): Customize › Connectors › Add custom connector, enter Flaghand and https://app.flaghand.com/mcp, then Add and Connect, and sign in. On Team and Enterprise an Owner adds it under Organization settings › Connectors, then each member connects.
• ChatGPT (Business, Enterprise and Edu, on the web): Plugins › + › Add custom MCP server, enter https://app.flaghand.com/mcp under Connection, choose OAuth, create it as a plugin, install it and sign in. OpenAI's help describes read access only on Pro and doesn't establish custom servers on Plus.
• Claude Code: an agent setting this up works over HTTP in its current session, with the steps below. For later sessions it registers the server with its token: claude mcp add --scope user --transport http flaghand https://app.flaghand.com/mcp --header 'Authorization: Bearer ${FLAGHAND_TOKEN}'. Only a person typing in Claude Code themselves runs claude mcp add --scope user --transport http flaghand https://app.flaghand.com/mcp, then /mcp to sign in.
• Codex: codex mcp add flaghand --url https://app.flaghand.com/mcp adds the server and signs in. With a saved workspace token instead: codex mcp add flaghand --url https://app.flaghand.com/mcp --bearer-token-env-var FLAGHAND_TOKEN (only the variable's name is saved).
• Cursor: merge {"mcpServers": {"flaghand": {"url": "https://app.flaghand.com/mcp"}}} into ~/.cursor/mcp.json, then connect Flaghand in Cursor's MCP settings and sign in.
• Gemini CLI: merge {"mcpServers": {"flaghand": {"httpUrl": "https://app.flaghand.com/mcp", "oauth": {"enabled": true}}}} into ~/.gemini/settings.json, then run /mcp auth flaghand in the next session and sign in.

In Claude and ChatGPT, new and existing accounts use the same sign-in page. Start here to get an account, or create it free on the sign-in page, then click Allow. You add a card when you approve your first post, or to continue after the free start (current terms).

A coding agent signing you up over HTTP

Claude Code, Codex or Gemini CLI with no Flaghand account yet:

1. Read the plans (no sign-in):

   curl https://app.flaghand.com/api/v1/plans
   

2. Start the sign-up:

   curl -X POST https://app.flaghand.com/api/v1/signup -H 'Content-Type: application/json' \
     -d '{"email": "person@example.com", "plan": "monthly", "agent_name": "Claude Code"}'
   

   Give the person signup_url and for_person as they are. Save claim_code to a private file such as claim.json: it's a secret, so keep it out of chat and command arguments.

3. Claim the token once they've finished:

   curl -X POST https://app.flaghand.com/api/v1/signup/claim -H 'Content-Type: application/json' -d @claim.json
   

   A 202 with "status": "pending" means they haven't finished: wait the Retry-After seconds and claim again. Put the token in your secret store and use it with Authorization: Bearer.

The full recipe is set up and connect.

Connect TikTok

You connect TikTok on TikTok's own sign-in and consent screen. Your TikTok password is never typed into Flaghand.

• In Flaghand: open Settings and choose Connect TikTok.
• From your agent: an agent that signed you up can start it with POST https://app.flaghand.com/api/v1/workspaces/{workspace}/connections/tiktok/start and give you the link. A connector token from Claude or ChatGPT can't start connections, so the agent sends you to Settings instead.

TikTok asks you to allow user.info.basic, video.publish and video.upload. The connection is pinned to that TikTok account. TikTok's login lasts 2 hours and renews on its own for up to 180 days. A login not renewed in 180 days, or one you revoked in TikTok, shows expired: connect it again.

Draft, approve, post

Ask your agent, for example: "Use Flaghand: make a five-slide TikTok photo post from these notes and send it to me for approval."

1. The agent drafts. A tiktok_photo draft has 1 to 35 photos (JPEG, PNG or WebP) or a slideshow Flaghand draws, a title of up to 90 characters and a description of up to 4,000. A tiktok_video draft has one MP4 or MOV of up to 64 MB and a title of up to 2,200 characters. post_mode is "direct" or "inbox".
2. The agent checks it. The check lists everything that would stop it posting, such as a video longer than TikTok allows the account. It also lists the choices that are yours to make.
3. The agent asks you. It sends one approval link. Nothing reaches TikTok before you approve.
4. You choose and approve. For a direct post, TikTok requires you to choose who can view it, with nothing chosen for you. Comments, and for videos Duet and Stitch, stay off unless you turn them on. You also say whether it's promotional content or a paid partnership. Then you approve with Face ID, Touch ID or a security key. Your passkey approval is your consent to TikTok's Music Usage Confirmation. An inbox post takes no choices here: you make them in TikTok.
5. Flaghand posts it. It goes out now or at its time. Just before a post at a time goes out, Flaghand reads TikTok's settings for the account again. If your choices no longer fit, nothing is sent, the post is held and you're told.

Afterwards the job reads posted with the post's link, inbox when it's waiting in your TikTok inbox, or failed with TikTok's reason. Flaghand never sends the same post to TikTok twice.

What the agent can and can't do on TikTok

The sign-in token has four scopes: read, drafts, media and request-approval. No scope can approve, schedule or post. Right before posting, Flaghand checks your signed approval again and asks TikTok for the account's settings again. If anything changed since you approved, nothing is sent and you approve again.

Other TikTok MCP servers

For jobs Flaghand doesn't do:

• TikTok for Business MCP Server is TikTok's own, for its ads platform. TikTok's help page says agents can create campaigns, pull performance reports, adjust budgets and refine audiences. It's about ads, not posting to a creator's account.
• Seym0n/tiktok-mcp reads TikTok through the TikNeuron service: a video's subtitles, a post's details and counts, and video search. It needs a TikNeuron API key. It doesn't post.

Questions

Is there a TikTok MCP for Claude? Yes. In Claude, open Customize › Connectors › Add custom connector, add https://app.flaghand.com/mcp and sign in. Claude can then draft TikTok photo and video posts and ask you to approve them. Claude Code, Codex, Cursor and Gemini CLI use the same server.

Is there an official TikTok MCP server? TikTok runs the TikTok for Business MCP Server, for its ads platform: campaigns, budgets, audiences and reports. Its help page doesn't describe posting to a creator's account. Flaghand posts through TikTok's official Content Posting API, after you sign in on TikTok.

Can my agent post to TikTok without asking me? No. Approving takes your passkey, and no Flaghand token can approve, schedule or post. Flaghand checks your signed approval again right before it posts.

Will my TikTok posts be public? Not yet. TikTok hasn't audited Flaghand's app, so a direct post works only for a private TikTok account, as Only me. Or Flaghand sends the post to your TikTok inbox as a draft, and you publish it in the TikTok app with the settings you choose there.

Can it read my TikTok analytics or other accounts? No. Flaghand reads only your account's name, picture and current posting options. You or your agent can record a post's views, likes and shares in Flaghand by hand.

How do I take the agent's access away? In Flaghand, open Settings › Agents and click Revoke. To disconnect TikTok, remove the account in Flaghand: it revokes Flaghand's access at TikTok and deletes the stored tokens.

Related guides

How to schedule TikTok posts compares TikTok Studio with tools and the API. TikTok's Content Posting API, explained covers Direct Post, the inbox and the audit. How to cross-post to X, LinkedIn and TikTok and video specs for TikTok, X and LinkedIn help with one post on several networks. Let an AI agent post safely explains the approval. Flaghand and TikTok covers permissions and what Flaghand keeps.