AI Agent Social Media Posting API: MCP, Scheduling, and Reposting

An AI agent can already plan a campaign, write a caption, choose a posting time, and prepare a video. The missing piece is a reliable action layer that can connect the agent to real social accounts.

Taisly Agent Kit provides that layer. It gives AI agents structured tools for discovering connected accounts, validating a video post, publishing immediately, scheduling for later, checking delivery status, and creating automatic repost rules.

The fastest option is the official Taisly remote MCP server:

https://app.taisly.com/mcp

It uses streamable HTTP and OAuth. For agents working with files on your computer, the same integration is available as a local MCP server, JSON-first CLI, JavaScript SDK, and direct social media posting API.

Short answer: If an AI agent can use MCP tools, run commands, or call an API, it can use Taisly to publish short-form video to connected social accounts. The safe workflow is always: authenticate, discover accounts, validate the post, confirm the final action with the user, publish or schedule, and save the returned historyId.

What is an AI agent social media posting API?

An AI agent social media posting API is an interface an agent can call to turn a high-level instruction into a real social media action. Instead of stopping after generating a caption, the agent can use structured tools to:

Check whether the social media connection is ready.
List the exact accounts available for publishing.
Read platform-specific posting constraints.
Connect a missing social account through a user-approved browser flow.
Validate a video URL or local video file.
Publish a video now or schedule it for a future time.
Return a stable post ID and check the result later.
Create and inspect cross-platform repost automations.

Taisly handles authentication with social platforms and the posting execution. The agent handles planning, content preparation, tool selection, validation, and orchestration. This separation lets developers add social publishing without rebuilding a separate TikTok, Instagram, YouTube, X, and Facebook integration for every agent.

What an AI agent can do with Taisly

User intent	Taisly capability	Result
"Which accounts can you use?"	List connected platforms	The agent receives exact account names, platform types, and IDs.
"Connect my Instagram account."	Start and check a browser connection	The user approves the account connection, then the agent verifies it.
"Is this video ready for TikTok?"	Read the platform schema and validate the post	Problems are found before a live publishing call.
"Post this launch video everywhere."	Create one post for selected account IDs	Taisly publishes the video to the explicitly selected destinations.
"Schedule this for Friday at 9:00 AM."	Add an ISO date and timezone to the post	The video is queued for the requested time.
"Did the post finish?"	Read recent post history or status by `historyId`	The agent reports the latest per-platform result.
"Repost future videos from YouTube to TikTok and Instagram."	Create a repost rule	New eligible source videos can be distributed automatically.

Taisly is focused on video publishing. It is not a video editor, a caption generator, or a replacement for an AI model. It is the execution layer that an agent can use after the media and publishing instructions are ready.

Four ways to connect an AI agent

There is no single setup that is best for every agent. The right choice depends mostly on where the video file lives and what tools the agent can use.

Connection method	Best for	Video input	Authentication
Taisly Application setup wizard	People who want guided setup for a known agent	Public URL or local file, depending on the selected integration	Guided OAuth or browser setup
Remote MCP	ChatGPT Apps, Claude connectors, hosted agents, and MCP clients	Direct public `http` or `https` video URL	OAuth, with API-key fallback for custom clients
Local MCP or CLI	Codex, Claude Code, Cursor, and agents that can read files on your computer	Local paths such as `./launch.mp4`	Browser setup credential or API key
SDK or REST API	Custom products, backend workers, and agent frameworks	File or URL according to the chosen endpoint	Taisly API key

1. Use the AI Agents Setup wizard in Taisly Application

The AI Agents Setup wizard is the easiest starting point for a human. It provides dedicated setup paths for:

Codex
ChatGPT
Claude
Claude Code
OpenClaw
Hermes Agent
Any other agent with MCP or terminal support

The wizard follows a simple Choose -> Connect -> Start flow. It generates the correct command, connector configuration, or copy-paste prompt for the selected agent. It also explains whether that setup expects a public video URL or can read a local video file.

Taisly AI Agents Setup wizard with Codex, ChatGPT, Claude, Claude Code, OpenClaw, Hermes Agent, and custom agent options — Choose an existing integration or select Other agent for a custom MCP or terminal-based setup.

After the user chooses an agent, Taisly displays only the relevant setup methods. For example, the Codex flow can generate one complete CLI setup prompt and also provides a separate remote MCP option.

Taisly Agent Kit showing a copyable CLI setup prompt for Codex and an alternative MCP tab — The Codex setup can be copied as one agent prompt, so Codex installs the CLI, returns the browser login URL, and verifies the connection.

2. Connect to the remote MCP server

Use remote MCP when the agent supports hosted MCP servers and the video is available through a public URL.

Server name: Taisly Social Media Posting
MCP endpoint: https://app.taisly.com/mcp
Transport: Streamable HTTP
Recommended authentication: OAuth
MCP Registry ID: io.github.taisly/agent

Remote MCP is a strong fit for ChatGPT Apps, Claude web connectors, Codex, Claude Code, Hermes Agent, and custom MCP clients. The user authorizes Taisly in the browser; the agent does not need to receive the user's social media passwords.

For Codex:

codex mcp add taisly --url https://app.taisly.com/mcp
codex mcp login taisly

For Claude Code:

claude mcp add --transport http taisly https://app.taisly.com/mcp

ChatGPT App setup

For ChatGPT, create a Taisly App or custom connector in a workspace that supports remote MCP tools. The Taisly wizard provides the Apps settings link, the connector name, the remote MCP URL, OAuth instructions, and the steps for enabling Taisly in a new chat.

Taisly AI Agents Setup instructions for creating a ChatGPT App with the Taisly remote MCP connector — ChatGPT connects to Taisly through Apps and Connectors using the official remote MCP endpoint and OAuth.

Claude connector setup

For Claude web, desktop, or mobile, add Taisly as a custom connector and use the same remote MCP URL. Taisly includes a direct connector link as well as manual instructions for individual and Team or Enterprise workspaces.

Taisly setup instructions for adding the official remote MCP server as a Claude custom connector — Claude can connect through the pre-filled Taisly connector link or through manual custom connector setup.

Read the complete MCP video publishing guide for connector setup and public URL requirements.

3. Run local MCP for local video files

A remote server cannot read ./video.mp4, a Downloads folder, or another private path on your computer. Use the local MCP server when the agent needs direct access to local media.

Start it with:

npx -y @taisly/agent mcp

Example MCP configuration:

{
  "mcpServers": {
    "taisly": {
      "command": "npx",
      "args": ["-y", "@taisly/agent", "mcp"]
    }
  }
}

Codex local MCP:

codex mcp add taisly -- npx -y @taisly/agent mcp

Claude Code local MCP:

claude mcp add --transport stdio taisly -- npx -y @taisly/agent mcp

Local MCP is also useful for editor agents such as Cursor, Windsurf, Cline, and Aider when their MCP configuration can launch a local command.

4. Use the CLI, SDK, or REST API

Install the Agent Kit CLI:

npm install -g @taisly/agent

Let the agent start a browser-based connection:

taisly setup --agent codex
# Open the returned loginUrl and approve the connection.
taisly checkin --agent codex
taisly auth:status

The browser setup stores a local agent credential. A manual API key from Taisly Settings is also available for backend workers and existing integrations.

Developers can use:

Remote MCP or local MCP: which one should an agent choose?

The decision is primarily about file access.

Situation	Recommended path
The video has a direct public URL such as `https://cdn.example.com/launch.mp4`	Remote MCP
The video is in the current project, Downloads folder, or another local path	Local MCP or CLI
The "URL" opens a cloud-storage preview page that requires login or cookies	Create a direct public download URL, or use local MCP
The workflow runs in a backend service	SDK or REST API
The user wants a guided setup without editing configuration	Taisly Application setup wizard

For remote MCP, a videoUrl must be reachable without cookies, a VPN, a login page, or an interactive confirmation screen. It must point to a downloadable video file, not a webpage that happens to contain a video. The current remote upload limit is 500MB.

For local MCP and CLI preflight, supported video extensions include .mp4, .mov, .avi, .mkv, .webm, .flv, .mpeg, and .mpg. Social networks may apply additional duration, codec, aspect-ratio, caption, privacy, or daily posting limits.

The MCP tools available to agents

The remote Taisly MCP server exposes structured tools with predictable JSON input and output.

Authentication and account discovery

taisly_auth_status checks the connection, plan limits, and connected account count.
taisly_platforms_list returns the social accounts the agent can use.
taisly_platform_schema returns posting constraints for a destination platform.
taisly_platform_connect_start creates a browser authorization link for a missing account.
taisly_platform_connect_check verifies that the account was connected.

Programmatic social-account connection currently supports the slugs instagram, tiktok, youtube, x, and facebook. The agent should always call taisly_platforms_list after connection and use the returned IDs instead of guessing.

Publishing, scheduling, and status

taisly_posts_validate checks a public video URL, destination IDs, caption, and optional schedule.
taisly_posts_create publishes now or schedules for later.
taisly_posts_status checks a post by historyId.
taisly_posts_list reads recent post history for status checks and audits.

taisly_posts_create requires confirmed: true. An agent should set it only after the user has approved the video, caption, destination accounts, and schedule.

Repost automation

taisly_reposts_list returns the current repost rules.
taisly_reposts_create creates an automation from one connected source account to one or more destination accounts.

The local MCP server also includes taisly_agent_setup_start and taisly_agent_checkin, which let a local agent guide the user through browser authentication without manually copying an API key.

The safe posting workflow for an AI agent

A reliable agent should not jump directly from a vague prompt to a live post. The recommended workflow is:

authenticate
-> list connected accounts
-> read platform requirements
-> validate the media and payload
-> show the final action to the user
-> receive explicit confirmation
-> publish or schedule
-> save historyId
-> check status

In practical terms:

Run taisly_auth_status.
Run taisly_platforms_list.
Match the user's requested destinations to exact platform IDs.
If an account is missing, use the connection tools and wait for browser approval.
Run taisly_platform_schema for every relevant destination.
Prepare the caption, video, settings, and scheduled time.
Run taisly_posts_validate.
Show the user the final video, caption, accounts, and schedule.
After approval, call taisly_posts_create with confirmed: true.
Return and store the historyId.
Use taisly_posts_status or taisly_posts_list to report the result.

This workflow gives the agent autonomy over the repetitive steps while keeping the consequential action under human control.

Example 1: publish a video now through remote MCP

Give an MCP-enabled agent a direct public video URL and an explicit request:

Use the Taisly MCP tools to publish this video:

Video URL: https://cdn.example.com/launch.mp4
Caption: Launch day. Here is the 30-second product tour.
Destinations: my TikTok and Instagram accounts
Schedule: publish now

First list my connected Taisly accounts and match the correct destinations.
Read the posting requirements and validate the post.
Show me the final video URL, caption, destination accounts, and timing.
Do not publish until I confirm.
After publishing, return the Taisly historyId and check the initial status.

The agent should first call read-only discovery and validation tools. After the user confirms, the final taisly_posts_create arguments can look like this:

{
  "videoUrl": "https://cdn.example.com/launch.mp4",
  "platforms": ["platform_id_1", "platform_id_2"],
  "description": "Launch day. Here is the 30-second product tour.",
  "confirmed": true
}

Example 2: schedule a social media post with an AI agent

Taisly accepts a scheduled date in the post payload. Include an explicit timezone so "9:00 AM" is not ambiguous.

Schedule ./launch.mp4 for Friday at 9:00 AM Asia/Bangkok.
Publish it to my connected TikTok, Instagram, YouTube, X, and Facebook accounts.
Use the caption: "A faster way to ship social video."

List the exact accounts first, validate the file and schedule, and show me the final action.
Wait for my confirmation before scheduling it.

Local CLI payload:

{
  "video": "./launch.mp4",
  "platforms": [
    "platform_id_1",
    "platform_id_2",
    "platform_id_3"
  ],
  "description": "A faster way to ship social video.",
  "scheduled": "2026-07-31T09:00:00+07:00"
}

Validate and create:

taisly posts:validate --json ./campaign.json
taisly posts:create --json ./campaign.json
taisly posts:status --id <historyId>

The agent should report the normalized scheduled time, destination accounts, and returned historyId.

Example 3: create a cross-platform repost automation

A repost rule is different from a one-time post. It tells Taisly to watch one connected source account and distribute eligible future videos to selected destination accounts.

Use Taisly to create a repost automation from my YouTube account to my TikTok and Instagram accounts.

List my existing repost rules and connected accounts first.
Show me the exact source and destination accounts.
Create the rule only after I approve it.

CLI example:

taisly reposts:list
taisly reposts:create \
  --from <youtube_platform_id> \
  --to <tiktok_platform_id>,<instagram_platform_id>

MCP agents use taisly_reposts_list and taisly_reposts_create for the same workflow.

For strategy and platform-specific details, see the complete video reposting guide, TikTok reposting guide, Instagram Reels reposting guide, and YouTube Shorts reposting guide.

Example 4: let the agent connect a missing social account

An agent does not need to stop permanently when the requested destination is missing.

The agent calls taisly_platforms_list.
It sees that the requested Instagram account is not connected.
It calls taisly_platform_connect_start with platform: "instagram".
Taisly returns a connectUrl.
The agent sends that URL to the user and waits.
The user signs in to Instagram and approves the connection in the browser.
The agent calls taisly_platform_connect_check.
It lists platforms again and continues only after the correct account appears.

The user remains in control of social-account authorization. The agent never needs the user's Instagram, TikTok, YouTube, X, or Facebook password.

You can also connect accounts manually with the Instagram connection guide and YouTube connection guide.

Which AI agents can use Taisly?

Taisly is not tied to one model vendor. The important requirement is tool access: the agent must be able to use remote MCP, launch a local MCP server, run the CLI, or call the API.

Chat and general-purpose agents

The application includes guided connection paths for ChatGPT and Claude. The same integration pattern can be used in workflows built with Gemini, Microsoft Copilot, Perplexity, Grok, DeepSeek, and Mistral AI when the surrounding environment can expose MCP tools or API calls.

Coding and computer-use agents

Taisly has dedicated workflow pages for Codex, Claude Code, OpenClaw, Hermes Agent, Cursor, Windsurf, Cline, Aider, GitHub Copilot Chat, Amazon Q, Replit Agent, Codeium, Tabnine, Bolt.new, Lovable, and v0.

These agents are especially useful when the video is part of a development workflow: a product demo, changelog clip, automated test recording, generated campaign asset, or launch announcement.

Agent frameworks and automation builders

Custom agent systems can wrap the Taisly tools inside larger workflows. Relevant examples include Manus, Lindy AI, Superagent, Vellum, Semantic Kernel, AutoGen, MetaGPT, and Agent Zero.

A typical multi-step workflow might generate a campaign brief, create or locate the video, write platform-specific copy, request human approval, call Taisly, and then monitor post status.

AI video and content tools

Video-generation and editing tools can sit upstream of the posting agent. Taisly provides workflow pages for tools such as Runway, Synthesia, HeyGen, Sora, Pika, Kling AI, InVideo, CapCut, Canva, Descript, and Opus Clip.

These products do not all expose the same tool interface. The safe general pattern is: one tool creates or edits the asset, an MCP- or API-capable agent obtains the final video file or URL, and Taisly performs the publishing action.

Browse the complete AI agent compatibility directory for every agent and content workflow page.

Which social platforms can an agent publish to?

Taisly is designed for short-form video distribution across connected social accounts, including:

The exact destinations available to an agent are always the accounts returned by taisly_platforms_list or taisly platforms:list. This is important because a platform name is not enough: the user may have multiple TikTok accounts, YouTube channels, Facebook pages, or Instagram profiles.

Never tell an agent to invent a platform ID or post to "all accounts" unless that is exactly what the user requested.

Platform-specific settings and captions

One video can be distributed to several platforms, but the ideal metadata may differ. An agent can use taisly_platform_schema before publishing to understand the destination's available settings.

Useful tasks for the agent include:

Rewriting the hook or call to action for each network.
Checking caption length and required fields.
Choosing privacy and platform-specific options.
Adding an explicit schedule with timezone.
Avoiding watermarks and low-quality downloaded copies.
Confirming that the vertical crop and duration fit the destination.

For content strategy, read how to post the same video without looking spammy, why reposted videos get low views, best times to post, and why social platforms compress videos.

Safety rules every publishing agent should follow

Publishing is an external action, so a good agent needs more than API access.

Discover accounts first. Never guess destination IDs.
Use the minimum requested scope. Do not post to every connected account unless the user explicitly asks.
Validate before creating. Local preflight catches common file and payload problems; Taisly performs final validation.
Show the final action. Confirm the media, caption, destination accounts, platform settings, and schedule.
Require explicit approval. Set MCP confirmed: true only after the user approves.
Protect credentials. Never print, log, or expose TAISLY_API_KEY.
Save the result. Keep the returned historyId for status and audit checks.
Do not retry blindly. If a create call times out, check Taisly History before sending the same post again.
Respect plan and platform limits. Explain the returned error instead of attempting to bypass it.
Respect platform rules. Taisly does not bypass account permissions, content policies, or media requirements.

JSON-first CLI workflow for local agents

The CLI is designed to return JSON so an agent can parse results without scraping human-oriented terminal output.

taisly auth:status
taisly platforms:list
taisly platforms:schema --platform TikTok
taisly posts:validate --json ./campaign.json
taisly posts:create --json ./campaign.json
taisly posts:list --page 1
taisly posts:status --id <historyId>
taisly reposts:list
taisly reposts:create --from <platform_id> --to <platform_id_1,platform_id_2>

integrations:list is also available as an alias for platforms:list. Taisly calls connected social accounts "platforms" in the application, while some APIs and agent frameworks call them "integrations."

JavaScript SDK example

Custom agents and backend services can use the package directly:

import { Taisly } from "@taisly/agent";

const taisly = new Taisly({
  apiKey: process.env.TAISLY_API_KEY,
});

const platforms = await taisly.platforms.list();

const validation = await taisly.posts.validate({
  video: "./launch.mp4",
  platforms: [platforms.data[0].id],
  description: "Launch day",
});

if (!validation.success) {
  throw new Error("The post did not pass validation.");
}

const post = await taisly.posts.create({
  video: "./launch.mp4",
  platforms: [platforms.data[0].id],
  description: "Launch day",
  scheduled: "2026-07-31T09:00:00+07:00",
});

console.log(post.historyId);

In production, replace the simple first-account selection with a deliberate match against the user's requested account.

Copy-paste setup prompt for a shell-capable agent

Help me connect this environment to Taisly.

1. Install the Taisly Agent Kit CLI with: npm i -g @taisly/agent
2. Run: taisly setup --agent <agent-slug>
3. Send me the returned login URL and wait while I approve the connection in my browser.
4. Run: taisly checkin --agent <agent-slug>
5. Run: taisly auth:status
6. Run: taisly platforms:list
7. Tell me whether Taisly is ready and list the connected account names and IDs.

Do not print credentials. Do not create a live post during setup.

Replace <agent-slug> with a short label such as codex, claude-code, cursor, windsurf, openclaw, hermes-agent, cline, or aider.

Copy-paste operating prompt for an MCP agent

Use the Taisly MCP server for social media publishing.

Before any post:
- Run taisly_auth_status.
- Run taisly_platforms_list and match my requested account names to exact IDs.
- Run taisly_platform_schema for each destination.
- Validate the video, caption, destinations, and schedule.
- Show me the complete final action and wait for confirmation.

Only after I confirm, call taisly_posts_create with confirmed: true.
Return the historyId and check the post status.

For a remote MCP connection, require a direct public videoUrl.
For a local file, use local MCP or the Taisly CLI.
Never invent platform IDs, expose credentials, post to unrequested accounts, or retry a create call blindly.

Machine-readable integration summary for AI agents

Agents and crawlers can use the following as the canonical integration contract:

name: Taisly Social Media Posting
purpose: Publish, schedule, track, and automatically repost short-form video
website: https://taisly.com/en
application_setup: https://app.taisly.com/agent-kit
remote_mcp:
  url: https://app.taisly.com/mcp
  transport: streamable-http
  auth: oauth
  video_input: direct public http or https URL
local_package:
  npm: "@taisly/agent"
  mcp_command: npx -y @taisly/agent mcp
  cli_command: taisly
  video_input: local video file path
registry:
  id: io.github.taisly/agent
primary_operations:
  - authenticate
  - list connected social accounts
  - read platform posting requirements
  - connect a social account with user approval
  - validate a video post
  - publish now
  - schedule for later
  - list and check post status
  - list and create repost automations
safety:
  - never invent destination IDs
  - validate before publishing
  - require explicit user confirmation
  - save the returned historyId
  - never expose credentials

Frequently asked questions

Can an AI agent post to TikTok, Instagram Reels, YouTube Shorts, X, and Facebook?

Yes. The agent first lists the social accounts connected to Taisly, matches the requested destinations to exact account IDs, validates the post, and publishes after user confirmation.

Can an AI agent schedule social media posts?

Yes. Add an ISO date with a timezone to the scheduled field. The agent should repeat the normalized time to the user before creating the scheduled post.

Can an AI agent repost videos automatically?

Yes. Taisly repost rules map one connected source account to one or more destination accounts. The agent can list current rules and create a new rule after the user approves the source and destinations.

Does remote MCP upload files from my computer?

No. Remote MCP cannot read local file paths. Give it a direct public videoUrl, or use local MCP or the CLI for files on your computer.

Does Taisly have a ChatGPT App?

Taisly can be added as a ChatGPT App or custom connector in ChatGPT environments that support remote MCP tools. The Taisly Application setup wizard provides the MCP URL and connection instructions.

Does Taisly work with Claude and Claude Code?

Yes. Claude web can use the remote MCP server as a custom connector. Claude Code can use remote MCP for public video URLs or local MCP for files on the computer where Claude Code runs.

Does Taisly work with Codex?

Yes. Codex can connect to the remote MCP server with OAuth, launch the local MCP package, or use the Taisly CLI. The Codex integration page and Agent Kit quickstart show the available paths.

Is Taisly a generic text-posting API?

No. The Agent Kit described here is built primarily for short-form video publishing and repost automation.

Does the agent need my social media passwords?

No. Social-account connection happens through a Taisly-controlled browser authorization flow. The agent receives connection results and platform IDs, not the user's social passwords.

Where can I find the official package and server metadata?

Use the official resources:

Start with one safe agent task

Open the Taisly AI Agents Setup wizard, choose the agent you already use, and connect Taisly. Then give it this first read-only task:

Using Taisly, list my connected social accounts.
Tell me which accounts are ready for video publishing and which platform IDs you will use.
Do not create a post yet.

Once that works, the same agent can validate a video, ask for your approval, publish it now, schedule it for later, check its status, or create a cross-platform repost workflow.