What Is an Agentic Social Media Scheduler?
Sources checked on October 12, 2026 · 56 sources
An agentic social media scheduler is a scheduling tool that an AI agent can operate for you. It still has a calendar and a composer, and it adds a way in for software: an MCP server, a command-line tool with an agent skill, or a REST API. You ask an assistant such as Claude or ChatGPT in plain words, the assistant calls the scheduler, and the post can be saved in the same calendar you would have used by hand, where that assistant’s documented limits allow the connection. “Agentic” is a label vendors use, us included. No standards body defines it.
What makes a scheduler “agentic”?#
The word comes from “agent”. Anthropic’s engineering guide defines the term: “Agents, on the other hand, are systems where LLMs dynamically direct their own processes and tool usage, maintaining control over how they accomplish tasks.”
So a scheduler is agentic, as this post uses the word, when its actions are offered to an agent as things it can call: list the connected accounts, create a post, read what is scheduled. The Model Context Protocol calls these tools, “Functions for the AI model to execute” (specification). A scheduler that only writes a caption when you click a button is not agentic in this sense, because you still do every step.
That definition is ours. The phrase itself is a vendor label: Postiz’s about page says “Postiz is the agentic social media scheduling platform”, and we use it for Postnavo too.
The terms, in plain words#
- AI agent. Software in which a language model decides its own next step and which tool to use.
- MCP (Model Context Protocol). In its own documentation, “an open-source standard for connecting AI applications to external systems”. Anthropic open-sourced it on November 25, 2024 and donated it to the Agentic AI Foundation, under the Linux Foundation, on December 9, 2025.
- MCP server, client and host. The server is the program that offers the tools, here the scheduler’s. The host is the AI application you type into, and the client is the part of it that holds the connection (architecture overview).
- Connector, app, plugin. Three names for a service connected to an assistant. Claude and Grok say “connectors”, ChatGPT says “apps”, bundled into “plugins”, and Gemini says “custom apps”.
- CLI. A command-line program: you, a script or a coding agent types a command and reads the text that comes back. One example, Postiz’s CLI, schedules posts “directly from your terminal or shell scripts”.
- Agent skill. “At its core, a skill is a folder containing a SKILL.md file” (Agent Skills). It tells an agent how and when to do a task, such as calling a CLI. The format is open and was first developed by Anthropic.
- REST API and webhook. An API is an interface for software, not for people (MDN). A REST API is, for a beginner, “an HTTP service that can be called using standard web libraries and tools” (MDN). A webhook runs the other way: the service sends data to your server when an event occurs (GitHub).
- OAuth. The sign-in window where you approve access without handing over a password. The standard “enables a third-party application to obtain limited access to an HTTP service”.
- Human in the loop. A person who can say no before the action happens.
MCP, CLI or API: three ways an agent gets in#
The table describes the three routes themselves, from the protocol’s and the format’s own documentation. It names no product.
| Question | MCP server | CLI with an agent skill | REST API with webhooks |
|---|---|---|---|
| What it is | A program that offers the scheduler’s actions to an AI application as tools | A command-line program, plus a skill: a folder with a SKILL.md file that says when and how to use it |
HTTP endpoints that any program can call, with no AI needed, and HTTP requests sent to your server when something happens |
| How the agent learns what it can do | It asks the server for its list of tools, each with a description and the inputs it accepts | It loads each skill’s name and description at start, and the full instructions when a task matches | Not documented as part of REST. Someone, or the agent, reads the API’s documentation |
| Where it runs | On the vendor’s platform, or as a process on your computer | On the machine where the agent runs | On the vendor’s servers. A webhook needs a URL of yours |
| Sign-in | Remote: based on OAuth 2.1, and optional. Local: credentials from the environment | Set by each tool, not standardised | Set by each API, not standardised by REST |
| Who is asked before a post is created | The protocol says a person should always be able to refuse a tool call, and that it cannot enforce this. The AI application shows the prompt | Depends on the coding agent and its permission mode. The skill format has only an optional, experimental field for tools “pre-approved to run” | Not documented: REST defines no approval step |
| Works with no chat window open | Not documented by the protocol: it depends on the AI application | Yes, where the agent has a non-interactive mode | Yes. Webhooks deliver events “as opposed to polling an API” |
| A good fit when | You chat with an assistant and want it to schedule what it writes | A coding agent or a release pipeline announces what it ships | An automation tool or your own product creates the posts |
Read from MCP’s architecture, tools and authorization pages, the Agent Skills overview and specification, Claude Code’s permissions and non-interactive mode, Codex’s CLI and approvals pages, Cursor’s CLI page, MDN and GitHub. The last row is our advice.
MCP server: for assistants you chat with#
Adding a remote server to a chat assistant comes down to entering its address and signing in, under each assistant’s own rules (next section). Postnavo’s MCP server is planned on this route.
CLI and agent skill: for coding agents#
Coding agents already run commands in a terminal. Claude Code “reads your codebase, edits files, runs commands”. A skill tells the agent that a command exists and when to use it. Postnavo’s page for this route is the CLI and agent skill.
REST API and webhooks: for automations and your own code#
This is the route when no assistant is involved: an automation tool, or your own product. A webhook tells your system when a post went out. Postiz’s documentation defines its own as “an HTTP callback Postiz fires when a post publishes”. Postnavo’s pages for this route are the REST API and webhooks and n8n, Make and Zapier workflows.
One tool can offer more than one route. Postiz’s MCP page says its MCP server “is built for AI assistants”, while its public API and CLI “are built for code and scripts”.
Which assistant uses which route#
Each row is what that assistant’s own documentation says. Each name leads to the page on how that assistant is planned to connect to Postnavo.
| Assistant | Route its documentation describes | Limits it states | Asks before a write? |
|---|---|---|---|
| ChatGPT | MCP: a custom MCP server, added on the web “as a plugin” | OpenAI’s pages do not state one rule (note 1). Plus, Go and Free: not documented | “Write actions by default require confirmation.” |
| Claude | MCP, as a “custom connector” (help page) | Free and every paid plan: “Free users are limited to one custom connector.” The server must be reachable over the public internet | Yes, with an “Allow always” option for tools “that you trust to run unsupervised” |
| Claude Code | MCP, skills and shell commands | None that matters here | Shell commands, in Manual mode: “Yes, except a built-in set of read-only commands” (permissions) |
| OpenClaw | Skills, on a gateway you run yourself | MCP: OpenClaw’s page and Postiz’s disagree (note 2) | “Commands run only when policy + allowlist + (optional) user approval all agree” (documentation) |
| Codex | A CLI that runs commands, MCP servers and skills | None that matters here | Set for each MCP server: “auto, prompt, writes, and approve”. The default value of default_tools_approval_mode is not stated. Tool calls with side effects: note 3 |
| Cursor | MCP, Agent Skills and a CLI | None that matters here | “Cursor asks for approval before using MCP tools by default.” |
| Gemini | Gemini app: a “custom app”, added in the web app by its MCP server URL (help page). Gemini CLI: MCP servers and Agent Skills | Gemini app: “Be 18 or over and in the US”, a personal Google Account, “Keep Activity on”, English only. Gemini CLI: “Unpaid tier and Google One users: Gemini CLI was replaced by Antigravity CLI on June 18th, 2026.” | Gemini app: “requires manual confirmation for any write actions”. Gemini CLI: yes, unless a server’s trust setting is on. Antigravity: “requiring your approval” by default (documentation) |
| Grok | MCP: “you can bring your own Model Context Protocol (MCP) server” (documentation) | “Connectors are available to all Grok users”. The server “must be reachable over the public internet” | Not documented |
- ChatGPT plans. OpenAI’s developer mode help page says: “Full MCP is only available to Business and Enterprise/Edu users, currently. Pro users can connect MCPs with read/fetch permissions in developer mode.” Its custom MCP server page and its Codex MCP page describe adding a server “as a plugin” and through “Add server” in the ChatGPT desktop app, and name no plan.
- OpenClaw and MCP. OpenClaw’s page, on saving “third-party MCP servers for OpenClaw-managed agent runs”: “OpenClaw is the MCP client-side registry and later projects those servers into eligible runtimes.” Postiz’s page: “No. OpenClaw runs skills and shell commands rather than MCP servers”. OpenClaw’s page is the authority on OpenClaw.
- Codex. OpenAI’s approvals page: “Codex can also elicit approval for app (connector) tool calls that advertise side effects, even when the action isn’t a shell command or file change. Destructive app/MCP tool calls always require approval when the tool advertises a destructive annotation (unless the tool advertises a read annotation, which takes priority).”
- Team plans. An admin or owner adds the server first. ChatGPT Business: “Only admins/owners can enable developer mode and deploy an app.” Claude: “only Owners and Primary Owners can add them to Team plans”. Grok Business and Enterprise: “a team admin must first provision a connector”.
The four chat assistants (ChatGPT, Claude, Gemini and Grok) each reach a scheduler through MCP, at a server address. This post gives no setup steps. Those belong to each assistant’s page, and every assistant Postnavo is launching with is listed in one place.
Why not copy and paste from ChatGPT?#
Copy and paste works, and it needs no setup. What it lacks is sight of your calendar: an assistant with no connection cannot see what is already scheduled. A connected one can, where the scheduler offers it: Buffer’s server lets an assistant “read your channels, browse your queue and drafts, schedule and edit posts” (developer documentation), and Postiz lists a tool to “List the organization’s posts scheduled between two dates” (documentation).
Whichever way you post, the character counter lists a character limit for each of 10 networks and checks a draft against them.
Who approves what the agent does?#
Can an agent publish something you have not seen? It depends on two gates, and neither is guaranteed.
The assistant asks before it acts#
MCP’s specification says: “there SHOULD always be a human in the loop with the ability to deny tool invocations.” It also says “MCP itself cannot enforce these security principles at the protocol level” (overview). So the prompt belongs to the assistant, and the table above gives what each one documents.
In several assistants the prompt can be turned off: Claude has “Allow always”, Gemini CLI has its trust setting, and ChatGPT has “Allow all actions, for an eligible app or account”, which OpenAI says “carries elevated risk because supported actions may run without another confirmation”. For a custom MCP server, OpenAI’s page describes a remembered choice that lasts one conversation: “New conversations will prompt for confirmation again.” The Gemini app’s page says confirmation is currently required for write actions: “Currently, Gemini requires manual confirmation for any write actions” (help page). Grok’s documentation does not say either way.
The scheduler holds the post#
This gate lives in the scheduler, so it differs by tool. Two tools, read from their own pages:
- Buffer. Its pages differ. The product page: “Everything goes through your Buffer queue, so nothing publishes without you seeing it first.” The developer documentation: “Leave the approval prompts on. create_post can publish immediately, and delete_post cannot be undone.”
- Postiz. The about page: “Every post an agent creates lands in your calendar first”. Its MCP tool list has a tool to “Schedule, draft, or immediately publish posts”, and its approvals page says “there is no “approved” flag that blocks publishing”. Read together, a draft is the only hold.
So ask the agent to save drafts, and keep the assistant’s prompt on. Postiz gives the drafting advice for its own built-in agent (documentation): “Ask it to draft rather than publish while you are getting a feel for it.” For a routine you can reuse, see an approval checklist for AI-drafted posts.
Schedulers that document agent access today#
Two schedulers whose documentation describes agent access:
- Postiz documents an MCP server, a CLI with a skill, a public API and webhooks. Start at docs.postiz.com.
- Buffer documents a remote MCP server, a CLI with skill files for Claude Code and Codex, and an API. Start at developers.buffer.com.
Both document agent access, so agent access alone does not set these two apart. This post does not compare approvals, networks or price.
The network’s rules still apply#
An agent does not change whose account it is, so read the network’s own rules before you connect one. Two of them, in the networks’ words.
X’s automation rules say: “You are ultimately responsible for the actions taken with your account, or by applications associated with your account.” Instagram’s Terms of Use rule out “accessing or collecting information in an automated way without our express permission”. For Instagram’s API rules, see what Instagram requires of any tool that posts for you. Other networks were not researched for this post.
FAQ#
Is an agentic scheduler the same as an AI social media agent?#
No. The agent is the part where a model decides what to do next, and the scheduler supplies the tools it calls. The agent can be one you bring, such as Claude or ChatGPT, or one built into the scheduler: Postiz describes its own as “a chat that can operate Postiz for you” (documentation).
Do I need to know how to code to use one?#
Not with a chat assistant. Google’s steps for Gemini include “Enter the app’s MCP server URL” (help page), and Buffer’s page answers the same question with “No. Most integrations connect with a single click.” The CLI and API routes assume you are comfortable in a terminal or an automation tool.
Do I need a paid ChatGPT, Claude, Gemini or Grok plan to connect a scheduler?#
It depends on the assistant: the table above gives what each one documents. Claude’s Free plan allows one custom connector. For ChatGPT, OpenAI’s developer mode help page says a custom server that can write is for Business, Enterprise and Edu, while its custom MCP server page and its Codex MCP page describe adding a server as a plugin and in the ChatGPT desktop app without naming a plan. A plugin from OpenAI’s directory “depends on your plan, workspace, role, region, and its included capabilities” (help page).
One vendor’s page names more plans than OpenAI’s help page. Postiz’s says: “Custom MCP connectors need ChatGPT Developer mode (Plus, Pro, Business, Enterprise, and Edu plans on the web).” OpenAI’s own pages do not state one rule, so check what your own account offers.
MCP or API: which one should I use?#
Neither is better. It depends on who makes the call: an assistant you chat with (MCP), a coding agent (a CLI and a skill), or code and automations (the API). The table above sets them side by side. It has no speed or cost figures, because none was found in a primary source.
Does the assistant get my social media passwords?#
Not through an OAuth sign-in. Your social accounts are connected to the scheduler, and the assistant is connected to the scheduler. Anthropic’s help page says the sign-in works “without Claude ever seeing your actual password”. What the connection can then reach is another matter: Buffer’s documentation says “The connection covers your whole account.”
Sources 56
ShowHide
Each page was read on the date beside it. The quoted sentence is the one this post relies on.
-
Model Context Protocol docs: What is the Model Context Protocol (MCP)? modelcontextprotocol.io
“MCP (Model Context Protocol) is an open-source standard for connecting AI applications to external systems.”
Checked
-
Model Context Protocol specification: overview (version 2026-07-28) modelcontextprotocol.io
“Tools: Functions for the AI model to execute [...] Hosts must obtain explicit user consent before invoking any tool [...] While MCP itself cannot enforce these security principles at the protocol level, implementors SHOULD:”
Checked
-
Model Context Protocol docs: Architecture overview modelcontextprotocol.io
“MCP Host: The AI application that coordinates and manages one or multiple MCP clients [...] MCP Client: A component that maintains a connection to an MCP server and obtains context from an MCP server for the MCP host to use [...] MCP Server: A program that provides context to MCP clients [...] MCP servers can execute locally or remotely. [...] MCP recommends using OAuth to obtain authentication tokens. [...] For example, a client can first list all available tools (tools/list) and then execute them. [...] description: Detailed explanation of what the tool does and when to use it”
Checked
-
Model Context Protocol specification: Tools modelcontextprotocol.io
“Each tool is uniquely identified by a name and includes metadata describing its schema. [...] the protocol itself does not mandate any specific user interaction model. [...] For trust & safety and security, there SHOULD always be a human in the loop with the ability to deny tool invocations.”
Checked
-
Model Context Protocol specification: Authorization modelcontextprotocol.io
“Authorization is OPTIONAL for MCP implementations. [...] Implementations using an STDIO transport SHOULD NOT follow this specification, and instead retrieve credentials from the environment. [...] A protected MCP server acts as an OAuth 2.1 resource server, capable of accepting and responding to protected resource requests using access tokens.”
Checked
-
Anthropic: Introducing the Model Context Protocol (dated November 25, 2024) anthropic.com
“Today, we're open-sourcing the Model Context Protocol (MCP), a new standard for connecting AI assistants to the systems where data lives, including content repositories, business tools, and development environments.”
Checked
-
Anthropic: Donating the Model Context Protocol and establishing the Agentic AI Foundation (dated December 9, 2025) anthropic.com
“Today, we’re donating the Model Context Protocol (MCP) to the Agentic AI Foundation (AAIF), a directed fund under the Linux Foundation, co-founded by Anthropic, Block and OpenAI, with support from Google, Microsoft, Amazon Web Services (AWS), Cloudflare, and Bloomberg.”
Checked
-
Anthropic engineering: Building Effective AI Agents anthropic.com
“Workflows are systems where LLMs and tools are orchestrated through predefined code paths. [...] Agents, on the other hand, are systems where LLMs dynamically direct their own processes and tool usage, maintaining control over how they accomplish tasks.”
Checked
-
Agent Skills: overview agentskills.io
“At its core, a skill is a folder containing a SKILL.md file. [...] Discovery: At startup, agents load only the name and description of each available skill, just enough to know when it might be relevant. [...] Activation: When a task matches a skill’s description, the agent reads the full SKILL.md instructions into context. [...] The Agent Skills format was originally developed by Anthropic, released as an open standard, and has been adopted by a growing number of agent products.”
Checked
-
Agent Skills: specification agentskills.io
“A skill is a directory containing, at minimum, a SKILL.md file: [...] The optional allowed-tools field: A space-separated string of tools that are pre-approved to run [...] Experimental. Support for this field may vary between agent implementations”
Checked
-
MDN Web Docs glossary: API developer.mozilla.org
“An API (Application Programming Interface) is a set of features and rules that exist inside a software program (the application) enabling interaction with it through software”
Checked
-
MDN Web Docs glossary: REST developer.mozilla.org
“Beginners can assume a REST API means an HTTP service that can be called using standard web libraries and tools.”
Checked
-
GitHub Docs: About webhooks docs.github.com
“Webhooks let you subscribe to events happening in a software system and automatically receive a delivery of data to your server whenever those events occur. [...] Webhooks are used to receive data as it happens, as opposed to polling an API (calling an API intermittently) to see if data is available. [...] When you create a webhook, you specify a URL and subscribe to events that occur on GitHub.”
Checked
-
IETF RFC 6749: The OAuth 2.0 Authorization Framework (abstract) datatracker.ietf.org
“The OAuth 2.0 authorization framework enables a third-party application to obtain limited access to an HTTP service”
Checked
-
Claude Help Center: Use connectors to extend Claude's capabilities support.claude.com
“Connectors let Claude access your apps and services, retrieve your data, and take actions within connected services. [...] You can also add custom connectors or connect to any service that supports MCP.”
Checked
-
Claude Help Center: Get started with custom connectors using remote MCP support.claude.com
“Custom connectors using remote MCP are available on Claude, Cowork, and Claude Desktop for users on Free, Pro, Max, Team, and Enterprise plans. Free users are limited to one custom connector. [...] This means your MCP server must be reachable over the public internet from Anthropic's IP ranges. [...] Note: While anyone can build and host connectors using remote MCP, only Owners and Primary Owners can add them to Team plans, and only Owners, Primary Owners, and people with a custom role that includes Manage access to Libraries can add them to Enterprise plans. [...] This allows Claude to interact with the application on your behalf, without Claude ever seeing your actual password. [...] Review Claude's tool approval requests carefully and only click "Allow always" when using a server and tool that you trust to run unsupervised.”
Checked
-
Claude Code docs: Overview code.claude.com
“Claude Code is an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools.”
Checked
-
Claude Code docs: Connect Claude Code to tools via MCP code.claude.com
“Claude Code can connect to hundreds of external tools and data sources through the Model Context Protocol (MCP), an open source standard for AI-tool integrations.”
Checked
-
Claude Code docs: Extend Claude with skills code.claude.com
“Claude Code skills follow the Agent Skills open standard, which works across multiple AI tools.”
Checked
-
Claude Code docs: Configure permissions code.claude.com
“Claude Code uses a tiered permission system to balance power and safety. The table shows, for each tool type, whether Manual mode asks before the action runs. [...] Bash commands [...] Shell execution [...] Yes, except a built-in set of read-only commands”
Checked
-
Claude Code docs: Run Claude Code programmatically code.claude.com
“Add the -p (or --print) flag to any claude command to run it non-interactively.”
Checked
-
OpenAI Help Center: Connected apps in ChatGPT help.openai.com
“Allow all actions, for an eligible app or account [...] Allow all actions carries elevated risk because supported actions may run without another confirmation. The standard account-wide and workspace-wide permission selectors do not offer this option. [...] An app connects ChatGPT to an external service, such as Google Drive or Slack, so it can access information or perform supported actions. A plugin packages capabilities into a workflow and can include one or more apps, skills, or both.”
Checked
-
OpenAI Help Center: Plugins in ChatGPT help.openai.com
“The plugin directory is available across ChatGPT plans. Installing or using an individual plugin depends on your plan, workspace, role, region, and its included capabilities.”
Checked
-
OpenAI Help Center: Developer mode and MCP apps in ChatGPT help.openai.com
“Note: Full MCP (Model Context Protocol) support, including modify/write actions, is rolling out in beta to ChatGPT Business, Enterprise, and Edu plans. [...] Only admins/owners can enable developer mode and deploy an app. [...] Full MCP is only available to Business and Enterprise/Edu users, currently. Pro users can connect MCPs with read/fetch permissions in developer mode.”
Checked
-
OpenAI developer documentation: Add custom MCP server developers.openai.com
“Connect your Model Context Protocol (MCP) server to ChatGPT as a plugin. ChatGPT supports both read and write tools from your server. [...] Access: Use ChatGPT on the web. [...] Find the resulting plugin in your personal plugins or the workspace where you created it. [...] Write actions by default require confirmation. [...] You can choose to remember the approve or deny choice for a given tool for a conversation, which means it will apply that choice for the rest of that conversation. [...] New conversations will prompt for confirmation again.”
Checked
-
OpenAI: Codex CLI learn.chatgpt.com
“Inspect code, make changes, run commands, and automate repeatable work without leaving your terminal. [...] Let Codex inspect files, make edits, and run the tools already installed on your machine.”
Checked
-
OpenAI: Model Context Protocol (Codex) learn.chatgpt.com
“The ChatGPT desktop app, Codex CLI, and IDE extension support MCP servers and share MCP configuration for the same Codex host. [...] Configure in the ChatGPT desktop app [...] Open Settings, then select MCP servers. [...] Select Add server. [...] default_tools_approval_mode (optional): Default approval behavior for tools from this server. Supported values are auto, prompt, writes, and approve.”
Checked
-
OpenAI: Build skills learn.chatgpt.com
“Use agent skills to extend ChatGPT and Codex with task-specific capabilities. [...] Skills build on the open agent skills standard.”
Checked
-
Cursor docs: Model Context Protocol (MCP) cursor.com
“Model Context Protocol (MCP) enables Cursor to connect to external tools and data sources. [...] Cursor asks for approval before using MCP tools by default.”
Checked
-
Cursor docs: Agent Skills cursor.com
“Agent Skills is an open standard for extending AI agents with specialized capabilities.”
Checked
-
Cursor docs: Cursor CLI cursor.com
“Cursor CLI lets you interact with AI agents directly from your terminal to write, review, and modify code. [...] Start a conversational session with the agent to describe your goals, review proposed changes, and approve commands: [...] Use print mode for non-interactive scenarios like scripts, CI pipelines, or automation:”
Checked
-
Gemini Apps Help: Connect & manage custom apps for Gemini Apps support.google.com
“To do this, add a custom app with its Model Context Protocol (MCP) server URL in the Gemini web app. [...] Be 18 or over and in the US. [...] For now, this feature isn’t available if you sign in with a work or school Google Account. [...] Have Keep Activity on. Custom apps are unavailable when this setting is off. [...] Only available in English. [...] Important: You can only connect custom apps in the Gemini web app. [...] Enter the app’s MCP server URL. [...] Currently, Gemini requires manual confirmation for any write actions, but Gemini can make mistakes.”
Checked
-
Gemini CLI docs: MCP servers with Gemini CLI geminicli.com
“Unpaid tier and Google One users: Gemini CLI was replaced by Antigravity CLI on June 18th, 2026. [...] An MCP server is an application that exposes tools and resources to the Gemini CLI through the Model Context Protocol, allowing it to interact with external systems and data sources. [...] trust (boolean): When true, bypasses all tool call confirmations for this server (default: false)”
Checked
-
Gemini CLI docs: Agent Skills geminicli.com
“Agent Skills let you extend Gemini CLI with specialized expertise, procedural workflows, and task-specific resources.”
Checked
-
Google Antigravity docs: MCP antigravity.google
“Antigravity supports the Model Context Protocol (MCP), an open standard that lets AI agents and editors securely connect to local developer tools, databases, file parsers, and external remote APIs. [...] By default, unconfigured MCP tools run in Ask mode, requiring your approval before execution.”
Checked
-
Grok docs: Connectors docs.x.ai
“Connectors are available to all Grok users and let Grok access your external tools and data sources directly within a conversation. [...] For Grok Business and Enterprise users, a team admin must first provision a connector in the cloud console before it is available to members of the organization. [...] If you need to connect Grok to a service not available in the catalog, you can bring your own Model Context Protocol (MCP) server. [...] Your MCP server must be reachable over the public internet.”
Checked
-
OpenClaw docs: home docs.openclaw.ai
“You run a single Gateway process on your own machine (or a server), and it becomes the bridge between your messaging apps and an always-available AI assistant.”
Checked
-
OpenClaw docs: Skills docs.openclaw.ai
“Skills are markdown instruction files that teach the agent how and when to use tools. [...] OpenClaw follows the AgentSkills spec.”
Checked
-
OpenClaw docs: MCP docs.openclaw.ai
“serve is OpenClaw acting as an MCP server. The other subcommands are OpenClaw acting as an MCP client-side registry for servers its own runtimes may consume later. [...] Save third-party MCP servers for OpenClaw-managed agent runs [...] OpenClaw is the MCP client-side registry and later projects those servers into eligible runtimes.”
Checked
-
OpenClaw docs: Exec approvals docs.openclaw.ai
“Commands run only when policy + allowlist + (optional) user approval all agree.”
Checked
-
Postiz: About Postiz postiz.com
“Postiz is the agentic social media scheduling platform. [...] Agents propose, people approve. Every post an agent creates lands in your calendar first, so you can review, edit or reschedule before anything goes live.”
Checked
-
Postiz documentation: MCP, Introduction docs.postiz.com
“The agent discovers these tools, understands their schemas, and calls them on your behalf. [...] Schedule, draft, or immediately publish posts [...] List the organization’s posts scheduled between two dates (with their current settings)”
Checked
-
Postiz documentation: CLI, Introduction docs.postiz.com
“For your AI agent to work best with Postiz, install the skill by running: [...] It wraps the Public API so you can schedule posts, manage integrations, and upload media directly from your terminal or shell scripts.”
Checked
-
Postiz documentation: API Overview docs.postiz.com
“Getting started with the Postiz Public API”
Checked
-
Postiz documentation: Webhooks docs.postiz.com
“A webhook is an HTTP callback Postiz fires when a post publishes.”
Checked
-
Postiz documentation: Sharing a post for approval docs.postiz.com
“Approval is a conversation, not a state: there is no "approved" flag that blocks publishing.”
Checked
-
Postiz documentation: The AI agent docs.postiz.com
“The agent is a chat that can operate Postiz for you. [...] Ask it to draft rather than publish while you are getting a feel for it. [...] The agent is Postiz’s own chat, using Postiz’s model.”
Checked
-
Postiz: Social Media MCP Server for AI Agents postiz.com
“Is OpenClaw an MCP client? [...] No. OpenClaw runs skills and shell commands rather than MCP servers, so it does not use the Postiz MCP link. [...] The MCP server is built for AI assistants: it exposes Postiz actions as tools the assistant calls from natural language and signs in with OAuth. The public API and the Postiz CLI are built for code and scripts, authenticate with an API key, and suit cron jobs, CI pipelines, and custom integrations.”
Checked
-
Postiz: Schedule Social Media Posts with ChatGPT postiz.com
“Custom MCP connectors need ChatGPT Developer mode (Plus, Pro, Business, Enterprise, and Edu plans on the web).”
Checked
-
Buffer: Buffer MCP Server buffer.com
“Your AI assistant drafts and schedules, and you approve. Everything goes through your Buffer queue, so nothing publishes without you seeing it first. [...] Do I need to know how to code? [...] No. Most integrations connect with a single click.”
Checked
-
Buffer developer documentation: MCP developers.buffer.com
“Buffer runs a remote MCP server at https://mcp.buffer.com/mcp, which enables AI assistants to read your channels, browse your queue and drafts, schedule and edit posts, capture ideas and pull post analytics, all without leaving the conversation. [...] The connection covers your whole account. [...] Leave the approval prompts on. create_post can publish immediately, and delete_post cannot be undone.”
Checked
-
Buffer developer documentation: home developers.buffer.com
“Connect Buffer to your agents, automation tools, or build something entirely new. [...] One API to build your own integrations for all major social platforms.”
Checked
-
X Help Center: Automation rules help.x.com
“You are ultimately responsible for the actions taken with your account, or by applications associated with your account.”
Checked
-
Instagram: Terms of Use help.instagram.com
“You can't attempt to create accounts or access or collect information in unauthorized ways. [...] This includes creating accounts or accessing or collecting information in an automated way without our express permission, regardless of whether such automated access or collection is undertaken while logged-in to an Instagram account.”
Checked
-
Buffer developer documentation: Buffer CLI developers.buffer.com
“Schedule posts, manage channels, and access account data directly from your terminal. Built for developers and AI agents. [...] The CLI ships with markdown skill files for Claude Code and Codex, allowing AI agents to interact with Buffer without requiring custom prompts or manual setup.”
Checked
-
OpenAI: Agent approvals & security (Codex) learn.chatgpt.com
“In the Auto preset (for example, --sandbox workspace-write --ask-for-approval on-request), Codex can read files, make edits, and run commands in the working directory automatically. [...] Codex asks for approval to edit files outside the workspace or to run commands that require network access. [...] Codex can also elicit approval for app (connector) tool calls that advertise side effects, even when the action isn’t a shell command or file change. Destructive app/MCP tool calls always require approval when the tool advertises a destructive annotation (unless the tool advertises a read annotation, which takes priority).”
Checked