Postproxy MCP Server

MCP (Model Context Protocol) server for integrating Postproxy API with Claude Code. This server provides tools for publishing posts, checking statuses, and managing social media profiles through Claude Code.

Installation

Global Installation

npm install -g postproxy-mcp

Local Installation

npm install postproxy-mcp

Claude Code stores MCP server configuration under ~/.claude/plugins/. After installing postproxy-mcp, Claude will automatically detect the server on restart.

Configuration

Register MCP Server

After installing postproxy-mcp, register it with Claude Code using the claude mcp add command:

claude mcp add --transport stdio postproxy-mcp --env POSTPROXY_API_KEY=your-api-key --env POSTPROXY_BASE_URL=https://api.postproxy.dev/api -- postproxy-mcp

Replace your-api-key with your actual Postproxy API key.

The configuration will be automatically saved to ~/.claude/plugins/. After running this command:

1. Restart your Claude Code session
2. Test the connection by asking Claude: "Check my Postproxy authentication status"
3. If tools are available, Claude will be able to use them automatically

Alternative: Interactive Setup

For non-technical users, you can use the interactive setup command:

postproxy-mcp setup

or

postproxy-mcp-setup

This will guide you through the setup process step by step and register the server using claude mcp add automatically.

Available Tools

Authentication Tools

auth_status

Check authentication status, API configuration, and workspace information.

Parameters: None

Returns:

{
  "authenticated": true,
  "base_url": "https://api.postproxy.dev/api",
  "profile_groups_count": 2
}

Account Overview

summary_get

Answer "what's the status?" in one call — an activity snapshot for a time window instead of separate history_list / comments_list / dm_chats_list round trips.

Parameters:

• window (string, optional): 24h (default), 7d, or 30d
• from (string, optional): ISO 8601 timestamp or bare date starting an explicit range. Overrides window, and the *_previous counts come back null
• to (string, optional): End of the explicit range. Defaults to now when only from is given
• profile_group_id (string, optional): Report on a single group. Omit to cover every group the key can reach

Returns:

{
  "window": {
    "label": "24h",
    "from": "2026-08-17T09:00:00Z",
    "to": "2026-08-18T09:00:00Z",
    "previous_from": "2026-08-16T09:00:00Z",
    "backlog_from": "2026-07-19T09:00:00Z"
  },
  "posts": {
    "published": 4,
    "published_previous": 3,
    "failed": 1,
    "scheduled_ahead": 6,
    "next_scheduled_at": "2026-08-18T14:00:00Z",
    "by_platform": { "instagram": { "published": 4, "failed": 0 } }
  },
  "engagement": {
    "total": { "impressions": 48210, "likes": 1204 },
    "by_platform": { "instagram": { "impressions": 31002, "likes": 900 } },
    "posts_with_insights": 14
  },
  "comments": { "received": 96, "received_previous": 71, "awaiting_reply": 12, "by_platform": { "instagram": 61 } },
  "reviews": { "received": 7, "received_previous": 4, "awaiting_reply": 3 },
  "dms": { "inbound": 41, "outbound": 33, "chats_awaiting_reply": 5, "reply_window_closing": 2 },
  "api": { "calls": 812, "calls_previous": 640 }
}

Notes:

• Post counts are posts, so a post sent to three networks counts once and a thread counts once. by_platform counts per-network deliveries, so a 3-item X thread is 3 under twitter.
• engagement is lifetime-to-date for posts published in the window, not engagement earned during it — it sums each post's newest stats snapshot. A post published minutes ago may have no snapshot yet and won't be in posts_with_insights. Keys are the normalized metrics listed in Stats Fields by Platform.
• The awaiting_reply counts describe current state, not the window — they don't change when you change window. They look back 30 days, returned as window.backlog_from. A comment counts as replied only when the reply came from you (via Postproxy or the profile itself); chats_awaiting_reply is derived from message timestamps, since Postproxy has no read/unread state.
• reply_window_closing counts chats with under 6 hours of their 24h messaging window left. Networks without a window (Telegram, Bluesky) are excluded.
• engagement is null when insights are off for the account; dms is null when DMs are off.
• Scoped like every other tool: a group-scoped key reports only its group.

Profile Management

profile_groups_list

List all profile groups accessible with your API key. Profile groups are organizational containers (e.g. per brand or client) that hold related profiles. Use a group's id to filter profiles_list by profile_group_id.

Parameters: None

Returns:

{
  "profile_groups": [
    {
      "id": "grp123abc",
      "name": "Main Brand",
      "profiles_count": 4
    }
  ]
}

profiles_list

List all available social media profiles for posting.

Parameters:

• profile_group_id (string, optional): If provided, only profiles in this group are returned (use profile_groups_list to find group IDs)

Returns:

{
  "profiles": [
    {
      "id": "profile-123",
      "name": "My Twitter Account",
      "platform": "twitter",
      "profile_group_id": "group-abc"
    }
  ]
}

profiles_placements

List available placements for a profile. For Facebook profiles, placements are business pages. For LinkedIn profiles, placements include the personal profile and organizations. For Pinterest profiles, placements are boards. For Telegram profiles, placements are channels the bot can post to. For Google Business profiles, placements are locations, returned as full resource paths (accounts/X/locations/Y) to pass as location_id. For WhatsApp profiles, placements are the phone numbers on the Business Account — the placement id is the phone_number_id every whatsapp_* tool and dm_chat_create take. Available for facebook, linkedin, pinterest, telegram, google_business, and whatsapp profiles.

Parameters:

• profile_id (string, required): Profile hashid

Returns (LinkedIn example):

{
  "placements": [
    {
      "id": null,
      "name": "Personal Profile"
    },
    {
      "id": "108520199",
      "name": "Acme Marketing"
    }
  ]
}

Notes:

• If no placement is specified when creating a post:
  • LinkedIn: defaults to the personal profile
  • Facebook: defaults to a random connected page (if only one page is connected, no need to set a placement ID)
  • Pinterest: it fails
  • Telegram: it fails — chat_id is required on every post
  • Google Business: it fails — location_id is required on every post and on every google_business_* tool

profiles_stats

Get the follower/engagement timeseries for a profile. Snapshots are captured roughly every 23 hours, so you can plot follower growth and other trends over time. The stats fields are platform-native (not normalized) — see Stats Fields by Platform in the post_stats section for shape and add-on profile-level keys (followers_count, followersCount, etc.) per network.

Parameters:

• profile_id (string, required): Profile hashid
• placement_id (string, conditional): Required for facebook, linkedin, telegram, and google_business profiles. Get it from profiles_placements. Omit for instagram, threads, youtube, twitter, tiktok, pinterest, and bluesky.
• from (string, optional): ISO 8601 timestamp — only include snapshots recorded at or after this time
• to (string, optional): ISO 8601 timestamp — only include snapshots recorded at or before this time

Returns (LinkedIn example):

{
  "data": {
    "profile_id": "prof_li_001",
    "platform": "linkedin",
    "placement_id": "108520199",
    "records": [
      { "stats": { "followerCount": 4500, "shareCount": 8, "likeCount": 80 }, "recorded_at": "2026-05-09T08:00:00Z" },
      { "stats": { "followerCount": 4520, "shareCount": 9, "likeCount": 90 }, "recorded_at": "2026-05-10T08:00:00Z" }
    ]
  }
}

For non-placement networks (e.g. Bluesky), omit placement_id:

{
  "data": {
    "profile_id": "prof_bsky_001",
    "platform": "bluesky",
    "placement_id": null,
    "records": [
      { "stats": { "followersCount": 8800, "postsCount": 40 }, "recorded_at": "2026-05-09T08:00:00Z" }
    ]
  }
}

Post Management

post_publish

Publish a post to specified social media profiles.

Parameters:

• content (string, required): Post content text

• profiles (string[], required): Array of profile IDs (hashids) or platform names (e.g., "linkedin", "instagram", "twitter"). When using platform names, posts to the first connected profile for that platform.

• schedule (string, optional): ISO 8601 scheduled time

• media (string[], optional): Array of media URLs or local file paths

• idempotency_key (string, optional): Idempotency key for deduplication

• require_confirmation (boolean, optional): If true, return summary without publishing

• draft (boolean, optional): If true, creates a draft post that won't publish automatically

• queue_id (string, optional): Queue ID to add the post to. The queue will automatically assign a timeslot. Do not use together with schedule.

• queue_priority (string, optional): Priority when adding to a queue: high, medium (default), or low

• platforms (object, optional): Platform-specific parameters. Key is platform name (e.g., "instagram", "youtube", "tiktok"), value is object with platform-specific options. See Platform Parameters Reference for full documentation.

  Example:

  json

  {
    "instagram": {
      "format": "reel",
      "collaborators": ["username1", "username2"],
      "first_comment": "Link in bio!"
    },
    "youtube": {
      "title": "My Video Title",
      "privacy_status": "public"
    },
    "tiktok": {
      "privacy_status": "PUBLIC_TO_EVERYONE",
      "auto_add_music": true
    }
  }
  

Returns:

{
  "post_id": "job-123",
  "accepted_at": "2024-01-01T12:00:00Z",
  "status": "pending",
  "draft": true
}

Note on draft posts: If you request a draft post (draft: true) but the API returns draft: false, a warning field will be included in the response indicating that the API may have ignored the draft parameter. This can happen if the API does not support drafts with certain parameters (e.g., media attachments) or under specific conditions. Check the warning field in the response for details.

post_status

Get status of a published post by job ID.

Parameters:

• post_id (string, required): Post ID from post.publish response

Returns:

{
  "post_id": "job-123",
  "overall_status": "complete",
  "draft": false,
  "status": "processed",
  "content": "Full post body as submitted...",
  "scheduled_at": "2024-01-02T09:00:00Z",
  "created_at": "2024-01-01T12:00:00Z",
  "source": "postproxy",
  "queue_id": null,
  "platforms": [
    {
      "platform": "twitter",
      "status": "published",
      "url": "https://twitter.com/status/123",
      "post_id": "123",
      "error": null,
      "attempted_at": "2024-01-01T12:00:00Z"
    }
  ]
}

scheduled_at is null for posts published immediately. Platform url is the published permalink (null until published).

Status values:

• overall_status: "draft", "pending", "processing", "complete", "failed"
• Platform status: "pending", "processing", "published", "failed", "deleted"
• Platform error: Error message if publishing failed (null if successful)

post_publish_draft

Publish a draft post. Only posts with draft: true status can be published using this endpoint.

Parameters:

• post_id (string, required): Post ID of the draft post to publish

Returns:

{
  "post_id": "job-123",
  "status": "processed",
  "draft": false,
  "scheduled_at": null,
  "created_at": "2024-01-01T12:00:00Z",
  "message": "Draft post published successfully"
}

post_delete

Delete a post by job ID.

Parameters:

• post_id (string, required): Post ID to delete

Returns:

{
  "post_id": "job-123",
  "deleted": true
}

post_stats

Get stats snapshots for one or more posts. Returns all matching snapshots so you can see trends over time. Supports filtering by profiles/networks and timespan.

Parameters:

• post_ids (string[], required): Array of post hashids (max 50)
• profiles (string, optional): Comma-separated list of profile hashids or network names (e.g. instagram,twitter or abc123,def456 or mixed)
• from (string, optional): ISO 8601 timestamp — only include snapshots recorded at or after this time
• to (string, optional): ISO 8601 timestamp — only include snapshots recorded at or before this time

Returns:

{
  "data": {
    "abc123": {
      "platforms": [
        {
          "profile_id": "prof_abc",
          "platform": "instagram",
          "records": [
            {
              "stats": {
                "impressions": 1200,
                "likes": 85,
                "comments": 12,
                "saved": 8
              },
              "recorded_at": "2026-02-20T12:00:00Z"
            }
          ]
        }
      ]
    }
  }
}

Stats fields by platform:

Notes: Instagram stories do not return stats. TikTok stats require the post to have a public ID.

Queue Management

queues_list

List all posting queues. Queues automatically schedule posts into recurring weekly timeslots with priority-based ordering.

Parameters:

• profile_group_id (string, optional): Filter queues by profile group

Returns:

{
  "queues": [
    {
      "id": "q1abc",
      "name": "Morning Posts",
      "description": "Daily morning content",
      "timezone": "America/New_York",
      "enabled": true,
      "jitter": 10,
      "profile_group_id": "pg123",
      "timeslots": ["Monday at 09:00 (id: 1)", "Wednesday at 09:00 (id: 2)"],
      "posts_count": 5
    }
  ]
}

queues_get

Get details of a single posting queue including its timeslots and post count.

Parameters:

• queue_id (string, required): Queue ID

queues_create

Create a new posting queue with weekly timeslots.

Parameters:

• profile_group_id (string, required): Profile group ID to connect the queue to (use profiles_list to find this)
• name (string, required): Queue name
• description (string, optional): Optional description
• timezone (string, optional): IANA timezone name (e.g. America/New_York). Default: UTC
• jitter (number, optional): Random offset in minutes (0–60) applied to scheduled times for natural posting patterns. Default: 0
• timeslots (array, optional): Initial weekly timeslots. Each object has day (0=Sunday through 6=Saturday) and time (24-hour HH:MM format)

Example:

{
  "profile_group_id": "pg123",
  "name": "Weekday Mornings",
  "timezone": "America/New_York",
  "jitter": 10,
  "timeslots": [
    { "day": 1, "time": "09:00" },
    { "day": 2, "time": "09:00" },
    { "day": 3, "time": "09:00" },
    { "day": 4, "time": "09:00" },
    { "day": 5, "time": "09:00" }
  ]
}

queues_update

Update a queue's settings, timeslots, or pause/unpause it. Changes to timezone or timeslots trigger rearrangement of all queued posts.

Parameters:

• queue_id (string, required): Queue ID to update
• name (string, optional): New queue name
• description (string, optional): New description
• timezone (string, optional): IANA timezone name
• enabled (boolean, optional): Set to false to pause the queue, true to unpause
• jitter (number, optional): Random offset in minutes (0–60)
• timeslots (array, optional): Timeslots to add or remove. To add: { "day": 1, "time": "09:00" }. To remove: { "id": 42, "_destroy": true }.

queues_delete

Delete a posting queue. Posts in the queue will have their queue reference removed but will not be deleted.

Parameters:

• queue_id (string, required): Queue ID to delete

queues_next_slot

Get the next available timeslot for a queue.

Parameters:

• queue_id (string, required): Queue ID

Returns:

{
  "next_slot": "2026-03-11T14:00:00Z"
}

Adding Posts to a Queue

When publishing a post with post_publish, you can add it to a queue instead of scheduling it manually:

• queue_id (string, optional): Queue ID to add the post to. The queue will automatically assign a timeslot. Do not use together with schedule.
• queue_priority (string, optional): Priority level: high, medium (default), or low. Higher priority posts get earlier timeslots.

Example:

{
  "content": "Queued post content",
  "profiles": ["twitter", "linkedin"],
  "queue_id": "q1abc",
  "queue_priority": "high"
}

Comment Management

comments_list

List comments on a published post. Returns paginated top-level comments with nested replies.

Parameters:

• post_id (string, required): Post ID
• profile_id (string, required): Profile ID to identify which platform's comments to retrieve
• page (number, optional): Page number, zero-indexed (default: 0)
• per_page (number, optional): Number of top-level comments per page (default: 20)
• from (string, optional): ISO 8601 date/time — only comments received at or after this point
• to (string, optional): ISO 8601 date/time — only comments received at or before this point

from/to filter on when Postproxy received the comment, not the platform's posted_at (which isn't always populated). A bare date such as 2026-03-25 means that date's start of day. The filter applies to top-level comments only — a comment in range still returns its full replies array.

Returns:

{
  "total": 42,
  "page": 0,
  "per_page": 20,
  "data": [
    {
      "id": "cmt_abc123",
      "external_id": "17858893269123456",
      "body": "Great post!",
      "status": "synced",
      "author_username": "someuser",
      "like_count": 3,
      "is_hidden": false,
      "posted_at": "2026-03-25T10:00:00.000Z",
      "replies": [
        {
          "id": "cmt_def456",
          "body": "Thanks!",
          "author_username": "author",
          "parent_external_id": "17858893269123456"
        }
      ]
    }
  ]
}

Comment objects may also include an attachments array (media on the comment — image, video, audio, gif, external, file), each with id, type, url, status, and external_id. Populated for Facebook, Threads, and Bluesky; Instagram, YouTube, and LinkedIn comments are text-only. The array is empty when there is no media.

comments_get

Get a single comment with its replies.

Parameters:

• post_id (string, required): Post ID
• comment_id (string, required): Comment ID (Postproxy ID or platform external ID)
• profile_id (string, required): Profile ID

comments_create

Create a comment or reply on a published post. The comment is published to the platform asynchronously.

Parameters:

• post_id (string, required): Post ID
• profile_id (string, required): Profile ID
• text (string, required): Comment text content
• parent_id (string, optional): ID of comment to reply to (Postproxy ID or external ID). Omit to comment on the post itself.

Returns:

{
  "id": "cmt_ghi789",
  "body": "Thanks for the feedback everyone!",
  "status": "pending",
  "external_id": null
}

The comment is created with status: "pending". Once published to the platform, it becomes "published". If publishing fails, it becomes "failed".

comments_delete

Delete a comment from the platform asynchronously. Supported on Instagram, Facebook, YouTube, and LinkedIn. Not supported on Threads.

Parameters:

• post_id (string, required): Post ID
• comment_id (string, required): Comment ID (Postproxy ID or external ID)
• profile_id (string, required): Profile ID

comments_edit

Edit the text of a comment you published, asynchronously. Supported on Facebook and YouTube only. Returns { accepted: true }; the outcome arrives via comment.edited / comment.edit_failed webhooks.

Parameters:

• post_id (string, required): Post ID
• comment_id (string, required): Comment ID (Postproxy ID or external ID)
• profile_id (string, required): Profile ID
• body (string, required): New comment text

comments_hide

Hide a comment on the platform asynchronously. Supported on Instagram, Facebook, and Threads.

Parameters:

• post_id (string, required): Post ID
• comment_id (string, required): Comment ID
• profile_id (string, required): Profile ID

comments_unhide

Unhide a previously hidden comment. Supported on Instagram, Facebook, and Threads.

Parameters:

• post_id (string, required): Post ID
• comment_id (string, required): Comment ID
• profile_id (string, required): Profile ID

comments_like

Like a comment on the platform asynchronously. Currently only supported on Facebook.

Parameters:

• post_id (string, required): Post ID
• comment_id (string, required): Comment ID
• profile_id (string, required): Profile ID

comments_unlike

Remove a like from a comment. Currently only supported on Facebook.

Parameters:

• post_id (string, required): Post ID
• comment_id (string, required): Comment ID
• profile_id (string, required): Profile ID

Platform Support

Direct Messages

1:1 messaging (chats and messages) on DM-capable profiles. Supported on Facebook (Messenger), Instagram (DMs), Telegram (Bot DMs), Bluesky, and WhatsApp. Outbound sends are processed asynchronously (returned with status: "pending"). Meta's 24h messaging window applies to Facebook/Instagram — a human replying to the participant's own inquiry can pass tag: "HUMAN_AGENT" to send outside it (up to 7 days, never for promotional or automated content); Telegram and Bluesky have no window. On WhatsApp the window is re-opened only by sending an approved template (template) or a category: "utility" text — see WhatsApp Business Management for the rest of the WhatsApp surface.

dm_chats_list

List chats for a profile, ordered by most recent activity.

Parameters:

• profile_id (string, required): Profile ID (Facebook, Instagram, Telegram, Bluesky, or WhatsApp)
• page (number, optional): Page number, zero-indexed (default: 0)
• per_page (number, optional): Items per page (default: 20)
• before / after (string, optional): ISO 8601 timestamp filters on last_message_at

WhatsApp chats also carry external_placement_id (the phone number the chat belongs to), group (true for group chats) and within_messaging_window.

dm_chat_create

Find or create a chat for a participant (idempotent — returns the existing chat if one exists). Use before messaging a participant the profile hasn't messaged yet.

Parameters:

• profile_id (string, required): Profile ID
• participant_external_id (string, required): Platform participant ID (IG-scoped user ID, Facebook PSID, Telegram user id, Bluesky DID, or a WhatsApp phone number with country code — non-digits are stripped)
• placement_id (string, optional): Placement id from profiles_placements. Required on WhatsApp (the phone number the chat is on); optional on Facebook (page id)
• participant_username (string, optional)
• participant_name (string, optional)

A brand-new WhatsApp conversation has no open messaging window, so the first send into it must be a template (see dm_message_send).

dm_chat_get

Get a single chat by Postproxy ID or platform external_conversation_id.

Parameters:

• chat_id (string, required): Chat ID or external conversation ID

dm_messages_list

List messages in a chat, most recent first.

Parameters:

• chat_id (string, required): Chat ID or external conversation ID
• page (number, optional): Page number, zero-indexed (default: 0)
• per_page (number, optional): Items per page (default: 20)
• direction (string, optional): inbound or outbound
• status (string, optional): Filter by message status

dm_message_send

Send an outbound message. Provide exactly one of body (text), media (a single attachment), or — on WhatsApp — template, interactive, location or contacts.

Parameters:

• chat_id (string, required): Chat ID or external conversation ID
• body (string, optional): Message text (required when nothing else is sent; on WhatsApp it may also accompany media as the caption)
• media (string[], optional): Up to one attachment as a URL or local file path. Not supported on Bluesky. (The remote/Worker MCP accepts URLs only.) WhatsApp limits: image 5MB jpeg/png, video 16MB mp4/3gpp, audio 16MB, document 100MB.
• tag (string, optional): HUMAN_AGENT to send outside the 24h window — extends it to 7 days from the participant's last inbound message (Facebook/Instagram only). Meta restricts it to a human replying to the participant's own inquiry; using it for marketing, offers, or automated re-engagement can get that Page / Instagram account's messaging capability suspended. Past 7 days Meta rejects the send and the message lands in status: failed with the platform error in error_details.
• reply_to_external_id (string, optional): Telegram & WhatsApp — platform message id (Telegram message_id, WhatsApp wamid) to quote / thread under
• reply_markup (object, optional): Telegram only — inline/reply keyboard payload
• quick_replies (object[], optional): Facebook & Instagram only — up to 13 tappable chips above the participant's composer. Each { title, payload }.
• buttons (object[], optional): Facebook & Instagram only — up to 3 buttons attached to the message. Each { type: "web_url", title, url } or { type: "postback", title, payload }.
• card (object, optional): Facebook & Instagram only — extra fields for the card carrying buttons (subtitle, image_url, default_action). Requires buttons.
• template (object, optional): WhatsApp only — send an approved template: { id | name, language, variables[], button_params[], header_media, header_location }. The only way to message outside the 24h window or to open a new conversation.
• interactive (object, optional): WhatsApp only — Meta-shaped interactive message (reply buttons, list, cta_url, product, flow, location request), passed through unchanged
• location (object, optional): WhatsApp only — { latitude, longitude, name, address }
• contacts (object[], optional): WhatsApp only — contact cards in Meta's contacts shape
• category (string, optional): WhatsApp only — "utility" marks a plain text send as a utility Direct Send that may go outside the window without a template
• link_preview (boolean, optional): WhatsApp only — false suppresses the URL preview on a text message
• voice_note (boolean, optional): WhatsApp only — deliver an OGG/Opus audio attachment as a voice note
• filename (string, optional): WhatsApp only — display name for a document attachment

Quick replies and buttons

Facebook Messenger and Instagram Direct only — on Telegram use reply_markup instead (passing these returns a 422). Quick replies are ephemeral chips that vanish once one is tapped; buttons stay attached to the message in the thread.

{
  "chat_id": "chat_xyz789",
  "body": "What can I help with?",
  "quick_replies": [
    { "title": "Track order", "payload": "TRACK" },
    { "title": "Talk to support", "payload": "HELP" }
  ]
}

{
  "chat_id": "chat_xyz789",
  "body": "Nike Air Max",
  "card": { "subtitle": "$129 · Arriving Friday", "image_url": "https://cdn.example.com/shoe.png" },
  "buttons": [
    { "type": "web_url", "title": "Buy now", "url": "https://shop.example.com/p/air-max" },
    { "type": "postback", "title": "Notify me", "payload": "NOTIFY:air-max" }
  ]
}

Buttons are delivered as a Meta generic template whose element title is your body — that's where the 80-character cap comes from. Instagram is stricter than Messenger: it delivers quick replies only on a plain-text message, so quick_replies with media or with buttons returns 422 there.

Receiving taps: a tapped chip or button postback arrives as an inbound message carrying tapped_action: { "kind": "quick_reply" | "postback" | "callback_query", "payload": "...", "title": "..." }. Read it from dm_messages_list / dm_message_get instead of digging through platform_data. Instagram ice-breaker taps and Telegram callback queries normalize to the same field.

WhatsApp templates and interactive messages

A WhatsApp number can only send free-form text, media, interactive, location and contacts while the participant's 24h window is open (within_messaging_window on the chat). Outside it — including a conversation you start yourself — send an approved template. variables fill the placeholders in order (header text vars first, then body vars, then dynamic URL button vars) and must match the template's variable_count from whatsapp_templates_list; the rendered body is stored on the message.

{
  "chat_id": "chat_wa1",
  "template": {
    "name": "order_update",
    "language": "en_US",
    "variables": ["Ana", "ORD-12345"]
  }
}

Templates with a media header take header_media: { "link": "https://..." }; a location header takes header_location; URL / copy-code / flow buttons take button_params: [{ "index": 0, "sub_type": "url", "parameters": [{ "type": "text", "text": "12345" }] }].

Inside the window, interactive is Meta's own object — reply buttons (up to 3), a list (up to 10 rows), a CTA URL, a product card, a flow or a location request:

{
  "chat_id": "chat_wa1",
  "interactive": {
    "type": "button",
    "body": { "text": "Ready to confirm your booking?" },
    "action": {
      "buttons": [
        { "type": "reply", "reply": { "id": "confirm", "title": "Confirm" } },
        { "type": "reply", "reply": { "id": "reschedule", "title": "Reschedule" } }
      ]
    }
  }
}

A tap comes back as an inbound message whose body is the tapped title, with the reply id under platform_data.interactive. Facebook/Instagram quick_replies / buttons return 422 on WhatsApp, and template / interactive / location / contacts / category return 422 on every other network.

dm_message_get

Get a single message by Postproxy ID or platform external_id.

Parameters:

• message_id (string, required): Message ID or external ID

dm_message_edit

Edit a previously-sent outbound message. Telegram only. Provide body and/or reply_markup (pass {} to clear the keyboard); at least one is required.

Parameters:

• message_id (string, required): Message ID or external ID
• body (string, optional): New text/caption
• reply_markup (object, optional): New keyboard (or {} to remove)

dm_message_react / dm_message_unreact

Add or remove your business account's reaction on a message. Facebook Messenger, Instagram Direct and WhatsApp.

Parameters (dm_message_react):

• message_id (string, required): Message ID or external ID
• reaction (string, optional): Named reaction (default love). Ignored on WhatsApp
• emoji (string, optional): Unicode emoji. Required on WhatsApp (any emoji)

Parameters (dm_message_unreact):

• message_id (string, required)

dm_chat_archive / dm_chat_unarchive

Archive (mute) or unarchive (unmute) a chat. Bluesky only. Returns the chat with archived set.

Parameters:

• chat_id (string, required): Chat ID or external conversation ID

dm_chat_mark_read

Mark a chat as read. On WhatsApp this sends the read receipt (blue ticks) for the newest inbound message; on other networks it only stamps metadata.read_at. Returns the chat.

Parameters:

• chat_id (string, required): Chat ID or external conversation ID

dm_comment_private_reply

Send a DM to the author of a comment, in reply to that comment (Meta "Private Replies"). Bypasses the 24h window (comments up to 7 days old) and creates/reuses a chat automatically. One private reply per comment, ever. Instagram and Facebook only.

Parameters:

• post_id (string, required): Post ID
• comment_id (string, required): Comment ID or external ID
• profile_id (string, required): Profile ID (Instagram or Facebook)
• text (string, required): DM text
• quick_replies (array, optional): Up to 13 chips — same shape as dm_message_send
• buttons (array, optional): Up to 3 buttons — same shape as dm_message_send; caps text at 80 characters
• card (object, optional): Card styling for buttons (subtitle, image_url, default_action)

Interactive elements follow the same rules as dm_message_send — on Instagram, quick_replies and buttons are mutually exclusive. Media attachments are not available on private replies.

Platform Support

History

history_list

List recent post jobs.

Parameters:

• limit (number, optional): Maximum number of jobs to return (default: 10)

Returns:

{
  "jobs": [
    {
      "post_id": "job-123",
      "content": "Full post body as submitted...",
      "content_preview": "Post content preview...",
      "created_at": "2024-01-01T12:00:00Z",
      "overall_status": "complete",
      "status": "processed",
      "scheduled_at": "2024-01-02T09:00:00Z",
      "draft": false,
      "source": "postproxy",
      "queue_id": null,
      "platforms_count": 2,
      "platforms": [
        {
          "platform": "twitter",
          "status": "published",
          "url": "https://x.com/user/status/123"
        }
      ]
    }
  ]
}

scheduled_at is null for posts that were published immediately. status is the raw API status (draft, scheduled, processing, processed, …), while overall_status folds platform outcomes into a single verdict.

Note: Postproxy's /posts API does not return profile identity (profile ID or name) per platform — only the network. Use profiles_list to map networks to connected profiles.

Google Business Profile Management

These edit the Google business listing itself — hours, attributes, services, food menus, action links and profile photos — as opposed to publishing local posts to it (that's post_publish with platform google_business).

Three rules apply to every tool in this group:

1. location_id is always required. It's the full Google resource path accounts/X/locations/Y, returned by profiles_placements.
2. Updates are field-masked. Each write takes a fields array naming exactly what's being replaced. Anything named in fields but absent from the payload is cleared, and nested objects are replaced wholesale rather than merged. Always read before you patch.
3. Availability varies by category and region. Attributes, service lists, food menus and action link types differ per listing. List what's available first; a location that isn't eligible returns 422.

Payloads use Google's own shapes and camelCased keys in both directions, so a response can be sent straight back as a request body.

Typical flow

1. profiles_placements                    → get location_id
2. google_business_location_get           → read current state
3. google_business_attributes_available   → see what this listing accepts
4. google_business_attributes_update      → patch only what changed

Attribute value shapes

google_business_attributes_available returns a valueType per attribute, which decides the shape to send back:

attribute_mask defaults to exactly the names you send, so a partial update never clears attributes you left out.

Hours

Times accept either "09:00" / "09:00:00" strings or Google's { "hours": 9, "minutes": 0 } objects — both are normalized before the call. Read current hours from google_business_location_get; each block named in fields is replaced entirely.

{
  "fields": ["regularHours"],
  "regularHours": {
    "periods": [
      { "openDay": "MONDAY", "openTime": "09:00", "closeDay": "MONDAY", "closeTime": "18:00" }
    ]
  }
}

Media

Google downloads the file from media_url itself — there is no upload step, so the URL must be publicly reachable https (not a short-lived signed URL, not localhost). Images need at least 250×250 and at most 5MB. category defaults to ADDITIONAL; use COVER or LOGO only when you intend to change the profile header or logo.

Reading empty results

Google omits keys rather than returning empty values: a listing with no attributes returns { "name": "..." } with no attributes key at all, and the same applies to serviceItems, placeActionLinks and boolean flags like isPreferred. Treat absent as empty.

Analytics

Google Business location analytics come through the standard profiles_stats tool — pass the location_id as placement_id. Metrics are Search and Maps impressions (desktop and mobile), website clicks, call clicks, direction requests, conversations, bookings, food orders and food-menu clicks. Google Business has no follower count, and Google exposes no per-post analytics for local posts.

WhatsApp Business Management

Manage a connected WhatsApp Business Account beyond the inbox: message templates, phone-number status and registration, the public business profile, display name and username, blocked users, groups, and Click-to-WhatsApp conversion reporting. Conversations themselves go through the dm_* tools above.

Four rules apply to every tool in this group:

1. The WhatsApp Business Account (WABA) is the profile; each phone number on it is a placement. Phone-scoped tools take phone_number_id — the placement id from profiles_placements (its metadata carries display_phone_number, quality_rating, messaging_limit_tier, name_status, platform_type).
2. Templates and the Conversions dataset are WABA-level — those tools take only profile_id.
3. Free-form messages only inside the 24h window. Once a participant's last inbound message is older than 24h (or the conversation is new), dm_message_send accepts only an APPROVED template (or a category: "utility" text). A delivered template re-opens the window.
4. Coexistence numbers are limited. A number connected with onboarding: "business_app" (still in the WhatsApp Business app on a phone) has lower throughput and cannot use the Groups API; whatsapp_number_info reports platform_type other than CLOUD_API for these.

Responses use Meta's own field names where Meta's shapes are returned (components, interactive, business profile fields).

Typical flow

1. profile_groups_initialize_connection   → platform: whatsapp (onboarding: api | business_app)
2. profiles_placements                    → get phone_number_id
3. whatsapp_templates_list                → find an APPROVED template and its variable_count
4. dm_chat_create                         → participant_external_id: phone, placement_id: phone_number_id
5. dm_message_send                        → template: { name, language, variables }
6. dm_messages_list                       → the reply opens a 24h window; free-form sends now work

Connecting a number

profile_groups_initialize_connection with platform: "whatsapp" returns a connect URL. Pass onboarding: "business_app" when the number lives in the WhatsApp Business app on a phone (the user scans a QR code; the app keeps working alongside Postproxy and up to 6 months of chats are imported) or onboarding: "api" when the number is with another provider or brand new (moved onto the Cloud API; no longer usable in the app). Omit it and the connect page asks. Each WhatsApp Business Account becomes one profile and each of its numbers a placement.

Templates

Meta reviews every custom template; a new one is PENDING until approved, and only APPROVED templates can be sent. Placeholders are {{1}}, {{2}}… (parameter_format: POSITIONAL, the default) or {{name}} (NAMED). variable_count on a template is the number of variables a send must carry: header text placeholders, then body placeholders, then one per dynamic-URL button. Library templates (whatsapp_template_library_get → whatsapp_template_create with library_template_name) skip review. A name is unique per language; deleting by name without language removes every language.

Conversions

Conversations that start from a Click-to-WhatsApp ad carry a ctwa_clid on the chat's metadata. Create a dataset once (whatsapp_dataset_create), then report outcomes with whatsapp_conversion_event_send by chat_id or phone; a chat with no captured ctwa_clid returns 422.

Example Prompts

Here are some example prompts you can use with Claude Code:

Check Authentication

Check my PostProxy authentication status

List Profiles

Show me all my available social media profiles

Publish a Post

Using profile IDs:

Publish this post: "Check out our new product!" to profiles ["profile-123"]

Using platform names:

Publish "Exciting news!" to linkedin and twitter

Publish with Platform Parameters

You can use platform-specific parameters to customize posts for each platform. The platforms parameter accepts an object where keys are platform names and values contain platform-specific options.

Instagram Examples

Regular Post with Collaborators:

Publish to Instagram: "Amazing content!" to my Instagram account with collaborators username1 and username2

Or with explicit parameters:

{
  "content": "Amazing content!",
  "profiles": ["instagram"],
  "media": ["https://example.com/image.jpg"],
  "platforms": {
    "instagram": {
      "format": "post",
      "collaborators": ["username1", "username2"],
      "first_comment": "What do you think? 🔥"
    }
  }
}

Instagram Reel:

{
  "content": "Check out this reel! #viral",
  "profiles": ["instagram"],
  "media": ["https://example.com/video.mp4"],
  "platforms": {
    "instagram": {
      "format": "reel",
      "collaborators": ["collaborator_username"],
      "cover_url": "https://example.com/thumbnail.jpg",
      "audio_name": "Trending Audio",
      "first_comment": "Link in bio!"
    }
  }
}

Instagram Story:

{
  "profiles": ["instagram"],
  "media": ["https://example.com/story-image.jpg"],
  "platforms": {
    "instagram": {
      "format": "story"
    }
  }
}

YouTube Examples

YouTube Video with Title and Privacy:

Upload this video to YouTube with title "My Tutorial" and make it public

Or with explicit parameters:

{
  "content": "This is the video description with links and details",
  "profiles": ["youtube"],
  "media": ["https://example.com/video.mp4"],
  "platforms": {
    "youtube": {
      "title": "My Tutorial: How to Build an API",
      "privacy_status": "public",
      "cover_url": "https://example.com/custom-thumbnail.jpg"
    }
  }
}

Unlisted YouTube Video:

{
  "content": "Video description",
  "profiles": ["youtube"],
  "media": ["https://example.com/video.mp4"],
  "platforms": {
    "youtube": {
      "title": "Private Tutorial",
      "privacy_status": "unlisted"
    }
  }
}

TikTok Examples

Public TikTok with Auto Music:

{
  "content": "Check this out! #fyp",
  "profiles": ["tiktok"],
  "media": ["https://example.com/video.mp4"],
  "platforms": {
    "tiktok": {
      "privacy_status": "PUBLIC_TO_EVERYONE",
      "auto_add_music": true,
      "disable_comment": false,
      "disable_duet": false,
      "disable_stitch": false
    }
  }
}

TikTok for Followers Only with AI Label:

{
  "content": "Special content for followers",
  "profiles": ["tiktok"],
  "media": ["https://example.com/video.mp4"],
  "platforms": {
    "tiktok": {
      "privacy_status": "FOLLOWER_OF_CREATOR",
      "made_with_ai": true,
      "brand_content_toggle": false
    }
  }
}

Facebook Examples

Facebook Post with First Comment:

{
  "content": "Check out our new product!",
  "profiles": ["facebook"],
  "media": ["https://example.com/product.jpg"],
  "platforms": {
    "facebook": {
      "format": "post",
      "first_comment": "Link to purchase: https://example.com/shop"
    }
  }
}

Facebook Story:

{
  "profiles": ["facebook"],
  "media": ["https://example.com/story-video.mp4"],
  "platforms": {
    "facebook": {
      "format": "story"
    }
  }
}

Facebook Page Post:

{
  "content": "Company announcement",
  "profiles": ["facebook"],
  "platforms": {
    "facebook": {
      "page_id": "123456789",
      "first_comment": "Visit our website for more details"
    }
  }
}

LinkedIn Examples

Personal LinkedIn Post:

{
  "content": "Excited to share my latest article on AI",
  "profiles": ["linkedin"],
  "media": ["https://example.com/article-cover.jpg"]
}

Company LinkedIn Post:

{
  "content": "We're hiring! Join our team",
  "profiles": ["linkedin"],
  "media": ["https://example.com/careers.jpg"],
  "platforms": {
    "linkedin": {
      "organization_id": "company-id-12345"
    }
  }
}

Bluesky Examples

Plain Bluesky post (auto-faceted mentions/tags/links):

{
  "content": "Hey @jay.bsky.team — check out our latest #ruby post: https://example.com/blog/post",
  "profiles": ["bluesky"]
}

You don't need any markup — Postproxy auto-converts @handles, #tags, and URLs into AT Protocol facets, and generates a link card preview from the URL's Open Graph meta (when no media is attached). 300-grapheme limit.

Telegram Examples

Telegram channel post (HTML formatting):

{
  "content": "<b>New release</b> — read more on our blog https://example.com/post",
  "profiles": ["telegram"],
  "platforms": {
    "telegram": {
      "chat_id": "-1001234567890",
      "parse_mode": "HTML",
      "disable_link_preview": true,
      "disable_notification": false
    }
  }
}

Use profiles_placements against your Telegram profile to list channel chat_ids the bot can post to. The bot must be added to the channel as administrator with permission to post.

Cross-Platform Examples

Same Content, Different Platforms:

{
  "content": "New product launch! 🚀",
  "profiles": ["instagram", "twitter", "linkedin"],
  "media": ["https://example.com/product.jpg"]
}

Video Across Platforms with Specific Parameters:

{
  "content": "Product launch video",
  "profiles": ["instagram", "youtube", "tiktok"],
  "media": ["https://example.com/video.mp4"],
  "platforms": {
    "instagram": {
      "format": "reel",
      "first_comment": "Link in bio!"
    },
    "youtube": {
      "title": "Product Launch 2024",
      "privacy_status": "public",
      "cover_url": "https://example.com/yt-thumbnail.jpg"
    },
    "tiktok": {
      "privacy_status": "PUBLIC_TO_EVERYONE",
      "auto_add_music": true
    }
  }
}

Platform Parameters Reference

Instagram:

• format: "post" | "reel" | "story"
• collaborators: Array of usernames (max 10 for posts, 3 for reels)
• first_comment: String - comment to add after posting
• cover_url: String - thumbnail URL for reels
• audio_name: String - audio track name for reels
• trial_strategy: "MANUAL" | "SS_PERFORMANCE" - trial strategy for reels
• thumb_offset: String - thumbnail offset in milliseconds for reels
• user_tags: Array of { username, x, y, media_index } - tag public Instagram accounts on any format (post, reel, story). Images require x and y (floats 0.0–1.0 from the top-left corner); reels and video slides are tagged by username only (coordinates are dropped); stories accept coordinates but don't need them. media_index picks the carousel slide (0-based, default 0). A leading @ is stripped. Out-of-range coordinates, a media_index past the last media item, or an image tag missing x/y are rejected with a 422 naming the entry. Private accounts and accounts with tagging off are silently skipped by Instagram.

YouTube:

• title: String - video title
• privacy_status: "public" | "unlisted" | "private"
• cover_url: String - custom thumbnail URL

TikTok:

• privacy_status: "PUBLIC_TO_EVERYONE" | "MUTUAL_FOLLOW_FRIENDS" | "FOLLOWER_OF_CREATOR" | "SELF_ONLY"
• photo_cover_index: Integer - index of photo to use as cover (0-based)
• auto_add_music: Boolean - enable automatic music
• made_with_ai: Boolean - mark content as AI-generated
• disable_comment: Boolean - disable comments
• disable_duet: Boolean - disable duets
• disable_stitch: Boolean - disable stitches
• brand_content_toggle: Boolean - mark as paid partnership (third-party)
• brand_organic_toggle: Boolean - mark as paid partnership (own brand)

Facebook:

• format: "post" | "story"
• first_comment: String - comment to add after posting
• page_id: String - page ID for posting to company pages

LinkedIn:

• organization_id: String - organization ID for company page posts

Telegram:

• chat_id: String, required — destination channel/chat ID (use profiles_placements to list)
• parse_mode: "HTML" | "MarkdownV2" — omit for plain text
• disable_link_preview: Boolean — suppress URL preview card
• disable_notification: Boolean — send silently (no notification sound)
• Character limit: 4,096 for text-only; 1,024 for the caption when media is attached (body beyond is truncated)
• Media: images ≤10 MB (×10), video ≤50 MB (×10), documents ≤50 MB (×1)
• The bot must be a member (preferably administrator with post permission) of the destination channel

Bluesky:

• No platform-specific parameters available
• Character limit: 300 graphemes (emoji and combining sequences count as one)
• Auto-detects @handle.bsky.social mentions, #hashtags, and URLs and converts them to clickable facets
• Generates a link card preview from Open Graph meta when a URL is present and no media is attached
• Media: images ≤1 MB (×4), video ≤100 MB (×1, 1–60s)
• Supports threads via the standard thread array

Twitter/X & Threads:

• No platform-specific parameters available

For complete documentation, see the Platform Parameters Reference.

Create a Draft Post

Create a draft post: "Review this before publishing" to linkedin

Publish a Draft Post

Publish draft post job-123

Check Post Status

What's the status of job job-123?

This will show detailed status including draft status, platform-specific errors, and publishing results.

Delete a Post

Delete post job-123

Get Post Stats

Show me the stats for post abc123

Get stats for posts abc123 and def456 filtered to Instagram only, from February 1st to today

List Placements

Show me the placements for my LinkedIn profile prof123

Queue Management

Show me all my posting queues

Create a queue called "Weekday Mornings" for profile group pg123, timezone America/New_York, with timeslots Monday through Friday at 9am

Add a post to queue q1abc with high priority: "Check out our latest feature!"

Pause queue q1abc

What's the next available slot for queue q1abc?

Comment Management

Show me the comments on post abc123 for my Instagram profile prof456

Reply to comment cmt_abc123 on post abc123 with "Thanks for the feedback!" using profile prof456

Hide comment cmt_abc123 on post abc123 for profile prof456

Direct Messages

List the DM chats for my Instagram profile prof456

Reply "Yes, we ship worldwide!" in chat chat_xyz789

Send a DM to the author of comment cmt_abc123 on post abc123 from profile prof456 saying "DM-ing you the details"

Open a WhatsApp chat with +1 310 555 0007 on number 1055512345 of profile prof_wa and send the order_update template in en_US with Ana and ORD-12345

Mark chat chat_wa1 as read and react to the last inbound message with 🔥

WhatsApp Business Management

List the approved WhatsApp templates on profile prof_wa

Create a UTILITY template called appointment_reminder in en_US on profile prof_wa: "Hi {{1}}, your appointment is on {{2}} at {{3}}."

Show the quality rating and messaging tier for number 1055512345 on profile prof_wa

Update the WhatsApp business profile of number 1055512345 on prof_wa: about "Open Mon–Sat 9–18", website https://acme.example

Block +1 310 555 0099 on number 1055512345 of profile prof_wa

Report a Purchase of 49.90 USD for chat chat_wa1 on profile prof_wa with event id ord-9921

View History

Show me the last 5 posts I published

Troubleshooting

Server Won't Start

• Check API Key: Ensure POSTPROXY_API_KEY is set when registering with claude mcp add
• Check Node Version: Requires Node.js >= 18.0.0
• Check Installation: Verify postproxy-mcp is installed and in PATH
• Check Registration: Ensure the server is registered via claude mcp add and configuration is saved in ~/.claude/plugins/

Authentication Errors

• AUTH_MISSING: API key is not configured. Make sure you included --env POSTPROXY_API_KEY=... when running claude mcp add
• AUTH_INVALID: API key is invalid. Verify your API key is correct.

Validation Errors

• TARGET_NOT_FOUND: One or more profile IDs don't exist. Use profiles_list to see available profiles.
• VALIDATION_ERROR: Post content or parameters are invalid. The API now returns detailed error messages:
  • 400 errors: {"status":400,"error":"Bad Request","message":"..."}
  • 422 errors: {"errors": ["Error 1", "Error 2"]} - Array of validation error messages
  • Check the error message for specific validation issues

API Errors

• API_ERROR: Postproxy API returned an error. Check the error message for details.
• Timeout: Request took longer than 30 seconds. Check your network connection and API status.

Platform Errors

When checking post status with post_status, platform-specific errors are now available in the error field of each platform object:

• error: null - Post published successfully
• error: "Error message" - Detailed error message from the platform API
• Common errors include authentication issues, rate limits, content violations, etc.

Draft Post Issues

If you create a draft post (draft: true) but receive draft: false in the response:

• The response will include a warning field explaining that the API may have ignored the draft parameter
• This can happen if:
  • The API does not support drafts with media attachments
  • The API has specific limitations for draft posts under certain conditions
• Check the warning field in the response for details
• Enable debug mode (POSTPROXY_MCP_DEBUG=1) to see detailed logging about draft parameter handling

Debug Mode

Enable debug logging by setting POSTPROXY_MCP_DEBUG=1 when registering the server:

claude mcp add --transport stdio postproxy-mcp --env POSTPROXY_API_KEY=your-api-key --env POSTPROXY_BASE_URL=https://api.postproxy.dev/api --env POSTPROXY_MCP_DEBUG=1 -- postproxy-mcp

Development

Building from Source

git clone https://github.com/postproxy/postproxy-mcp
cd postproxy-mcp
npm install
npm run build

Running in Development Mode

npm run dev

License

MIT