Your AI agents post.
You approve.
A REST API and MCP server that lets your AI agents post to social media, with a human approval queue before anything goes live.
Works with ChatGPT, Claude, Cursor, n8n and any MCP client
Live demo
Watch an agent post, with you in the loop.
The agent drafts and schedules. Nothing goes live until a person approves it.
What is Postgale?
Postgale is the secure layer for AI agents on social media: a REST API and MCP server that lets businesses, agencies and developers connect ChatGPT, Claude or their own agent to Instagram, Facebook, Threads, TikTok, WhatsApp, Telegram, Discord, Bluesky and Mastodon, with a human approval queue before anything goes live.
Building with an AI agent? Connect it over MCP or the REST API.
Posts, replies and messages on the networks your audience uses
How it works
How does Postgale work? Five steps.
Your agent talks to Postgale instead of straight to the networks. Postgale holds what needs a person, publishes the rest, and writes everything down.
Connect your social accounts
Sign in to Instagram, Facebook, Threads, TikTok and the rest from the Postgale dashboard with each network’s own OAuth screen (Telegram uses a one-time code, Bluesky an app password). Postgale stores the tokens encrypted and refreshes them; the agent never sees a password.
Connect your agent
Add the MCP server URL https://postgale.com/mcp to ChatGPT, Claude, Cursor or VS Code and sign in with OAuth, or create an API key (sk_…) for n8n, your own code and the REST API at https://postgale.com/api/v1.
The agent creates the post
The agent calls the create_post tool, or POST /api/v1/posts, with the text, media, accounts and an optional time. Where approval is on (for the workspace, an account, an API key or an AI app) the post is saved with the status pending_approval instead of going out. Direct TikTok posts from an agent always wait.
A person approves it in the queue
A signed-in member of the workspace approves, edits or rejects it in the dashboard’s approval queue. API keys can’t approve: the approve endpoint answers 403 and the MCP server has no approve tool. Comment replies, DMs and edits to live posts wait in the same queue.
Postgale publishes and records it
The approved post goes live now or at its scheduled time. Your agent gets a signed post.published (or post.failed) webhook, get_post shows the status and URL on each network, and every write is in the audit log with the key or app that made it.
Setup
Connect in 2 minutes.
Nothing to install. Paste one URL into your AI tool, or one header into your HTTP calls.
Using n8n, OpenClaw or Hermes Agent? See every setup guide · MCP reference · OpenAPI 3.1
MCP + REST
One MCP server, one REST API. Every network.
The same actions as MCP tools for AI assistants and as REST endpoints for your own code, on the networks below.
Which social networks are supported?
- InstagramPhotos, Reels, Stories and carousels; comments, mentions and DMs (professional accounts)
- Facebook PagesText, link, photo, video, Reel and Story posts; comments and Messenger
- ThreadsText, image, video and carousel posts, reply chains; replies
- TikTokVideo and photo posts; a person picks privacy settings before an agent’s post goes out
- WhatsApp BusinessMessages, media and templates (messaging, not posts)
- TelegramChannel and group posts, albums; edit and delete
- DiscordMessages, embeds, polls, forum posts and threads; roles, pins, events and DMs
- BlueskyPosts, threads, comments and text DMs with images, video and link cards
- MastodonStatuses on any server with media, alt text and content warnings
- Coming soon not available yet
Which MCP tools and REST endpoints does an agent call?
| Task | MCP tool | REST endpoint |
|---|---|---|
| List connected accounts | list_accounts | GET /api/v1/accounts |
| Get an OAuth link to connect an account | connect_account | GET /api/v1/connect/{platform} |
| Upload an image or video | upload_media | POST /api/v1/media |
| Create, schedule or publish a post | create_post | POST /api/v1/posts |
| Read a post’s status on each network | get_post | GET /api/v1/posts/{id} |
| Reply to a comment | reply_to_comment | POST /api/v1/comments/{id}/reply |
| Send a direct message | send_message | POST /api/v1/inbox/conversations/{id}/messages |
| List actions waiting for approval | list_actions | GET /api/v1/actions |
| Register a webhook | create_webhook | POST /api/v1/webhooks |
| Approve a postNo MCP tool; answers 403 to API keys (dashboard only) | none | POST /api/v1/posts/{id}/approve |
How does an agent authenticate?
- MCP: https://postgale.com/mcp (Streamable HTTP, 132 tools). Sign in with OAuth 2.1 from ChatGPT, Claude, Cursor or VS Code, or send Authorization: Bearer sk_… from any other client.
- REST: https://postgale.com/api/v1, JSON in and out, Authorization: Bearer sk_… with an API key from the dashboard. Each key acts for one workspace and can be revoked; sign-in tokens work on the MCP server only.
Security
Built so the agent can’t go around you.
The controls live in Postgale, not in the prompt. An agent can’t switch them off, and neither can a leaked key.
Agents can’t approve their own work
Approving happens only in the dashboard, by a signed-in person. The API answers 403 to a key that tries, and the MCP server has no approve tool.
Approval exactly where you want it
Per workspace, account, API key or connected AI app. Comment replies, DMs, moderation and edits to live posts wait in the same queue as posts.
Edits go back to review
People approve the exact version they saw. If the text changes after approval, the post returns to the queue.
An audit log of every write
Each action records whether a person, an API key or a connected app made it, and who approved it.
Your tokens never reach the agent
Platform tokens are encrypted with AES-256-GCM and never returned by the API. The agent holds only a Postgale key, stored as a SHA-256 hash and shown once.
Revoke in one click
Revoke a key or a connected app from the dashboard and its next request fails. An approved action whose key was revoked doesn’t run.
Approval events (post.approved, post.rejected, action.executed and more) reach your webhooks signed with HMAC-SHA256.
Use cases
The AI social media manager you wanted, with a brake.
Let the agent do the daily work. Keep a person on what goes out.
Safe AI agents on social media
The controls that let an agent post, reply and message without being able to go around a person.
Read the use caseAI social media manager
An agent that drafts, schedules, watches comments and answers DMs, with a person approving where it matters.
Read the use caseAgency client approvals
A workspace per client, agents drafting, the team approving in one queue.
Read the use caseWhatsApp and Instagram agent for a small business
An agent that answers customers on WhatsApp and Instagram and posts the week’s updates.
Read the use caseChatGPT social media automation
ChatGPT drafts, schedules and answers through a Postgale connector, under your approval.
Read the use case
Compare
How does Postgale compare with other social media APIs?
Built only from each vendor’s own public pages, with sources. Some cover more networks or publish prices today; we say so.
| Product | Human approval | MCP server | Networks | Pricing |
|---|---|---|---|---|
| Postgale | Set by a person per workspace, account, API key or AI app; keys can’t approve; replies and DMs queue too | Hosted, 132 tools, OAuth sign-in or API key | 9 live; X and LinkedIn coming soon | Not published yet |
| Postgale vs Ayrshare | A requiresApproval flag on the post request; approved through its API | Listed on its site | 14 | From $149/month |
| Postgale vs Buffer API | Approval workflows between teammates (Team plan) | A “Social Media MCP” page; integrations for ChatGPT, Claude and Cursor | 12 channels | Free for 3 channels; from $5/channel/month (yearly) |
| Postgale vs Upload-Post | Not described on the pages checked | 50 tools with an API key; open source | 22 | Free tier; from $16/month (yearly) |
Vendor facts as checked on 22 October 2026. More: Postgale vs Postiz · Postgale vs Outstand · Postgale vs Blotato · Postgale vs Composio · Postgale vs PostEverywhere · all comparisons
Updated
Postgale is the secure layer for AI agents on social media: a REST API and MCP server that lets businesses, agencies and developers connect ChatGPT, Claude or their own agent to Instagram, Facebook, Threads, TikTok, WhatsApp, Telegram, Discord, Bluesky and Mastodon, with a human approval queue before anything goes live.
Your agent connects to Postgale instead of straight to the networks. Postgale keeps the account logins, applies each platform’s rules, holds what needs a person, publishes and records every action. Developers get a REST API and a hosted MCP server with 132 tools; everyone else gets a dashboard with the approval queue, posts, comments and messages. It is operated by Logic Minds LLC.
Connect ChatGPT to Postgale’s MCP server and your Instagram professional account to Postgale; ChatGPT can then publish and schedule photos, Reels, Stories and carousels. In ChatGPT, turn on developer mode, create a connector with the URL postgale.com/mcp and OAuth authentication, sign in and pick your workspace.
When you allow the connection you can tick “Hold posts from this app for approval”, so what ChatGPT writes waits in your queue until a person approves it. Personal Instagram accounts can’t publish through Meta’s API; Business and Creator accounts can.
For Threads and TikTok, add postgale.com/mcp as a custom connector in claude.ai, Claude Desktop or the mobile apps (or run claude mcp add in Claude Code), sign in and connect the accounts in Postgale. LinkedIn is not supported by Postgale today.
Claude can then post text, images, video and carousels to Threads, and videos and photo posts to TikTok. TikTok requires a person to choose each post’s privacy and interaction settings, so direct TikTok posts from Claude always wait in the approval queue (inbox drafts go to the creator’s TikTok app instead).
Yes: Postgale hosts one at postgale.com/mcp, with 132 tools for accounts, media, posts, comments, direct messages, WhatsApp, TikTok, analytics and webhooks. Where approval is on, create_post saves the post as pending_approval and a person approves it in the dashboard; there is no approve tool.
MCP (Model Context Protocol) is the open standard that lets AI assistants such as ChatGPT, Claude and Cursor use outside tools. Each Postgale tool is labelled read-only, write or destructive, so a client can ask you before anything goes live.
Yes, if the controls sit outside the agent: with Postgale the agent never gets your passwords or platform tokens, a person approves what goes live, and every action is logged. The agent holds only a Postgale API key or sign-in that you can revoke at any time.
Platform tokens are encrypted at rest with AES-256-GCM; API keys are shown once and stored as SHA-256 hashes; each key acts for one workspace. Every write through the API or MCP is recorded in the audit log with the key or app that made it, and webhooks are signed with HMAC-SHA256.
It depends on what you need most. If a person must approve what the agent publishes, and replies and DMs should wait too, Postgale is built for that: an MCP server and REST API with an approval queue the agent can’t skip and an audit log.
If you need LinkedIn, YouTube, Pinterest or X today, or published prices with a free tier, broader APIs such as Ayrshare or Upload-Post cover more networks. The comparison pages lay out the facts side by side, with sources on each vendor’s own site.
Ayrshare is a broad social media API: it covers 14 networks, from $149 a month. Postgale is narrower and built around control of the agent.
Ayrshare’s approval is an opt-in requiresApproval flag on each post, approved through its API. In Postgale a person sets approval in the dashboard, the agent’s key can’t approve, and replies and DMs wait in the same queue. Postgale has not published prices yet.
No. Approving and rejecting only happen in the Postgale dashboard, by a signed-in member of the workspace. The approve and reject endpoints answer 403 to API keys, and the MCP server has no approve tool. If a post is edited after approval, it goes back to the queue.
Yes: you turn approval on for the whole workspace, for one social account, for one API key or for one connected AI app. For example, a scheduler running from n8n can post on its own while everything ChatGPT writes waits for you. Direct TikTok posts from an agent always wait for a person.
When approval applies, posts wait as pending_approval, and comment replies, moderation, direct messages, WhatsApp messages and changes to published posts wait in the same queue. Rules that decide by content (for example, only posts that mention prices) are not available yet.
Publishing on Instagram (photos, Reels, Stories, carousels), Facebook Pages, Threads, TikTok, Bluesky, Mastodon, Telegram channels and groups, and Discord channels. Messaging on WhatsApp Business, Instagram DMs and Messenger. Comments and mentions are read and answered on Instagram, Facebook Pages and Threads.
X and LinkedIn are coming soon and cannot be connected yet. YouTube and Pinterest are not supported.
ChatGPT (as a connector in developer mode), Claude (claude.ai, Claude Desktop, the mobile apps and Claude Code), Cursor, VS Code and any other MCP client connect to postgale.com/mcp and sign in with OAuth. n8n, OpenClaw, Hermes Agent and your own code use the REST API or the MCP server with an API key.
Every connection shows up in the dashboard, where you can require approval for it or revoke it.
Postgale is invite-only for now: join the waitlist on this page and we set each team up on a short call. Prices are not published yet. Questions: [email protected].
Waitlist
Give your agent the work. Keep the final say.
We’re onboarding a few teams each week. Tell us how you work and we’ll set you up personally.
- A short setup call: we connect your accounts and your agent with you
- Approval on from the first post, set where you want it
- Works with ChatGPT, Claude, Cursor, n8n or your own code
Already have access? Log in · Questions? [email protected]