feat: Tumblr platform + platform availability (node 1.6.0, mcp 1.9.0, python 0.7.0)

Syncs the SDKs and docs to the webapp contract changes.

Tumblr (15th platform):
- `tumblr` added to every platform enum — openapi Channel + PlatformAvailability
  schemas, node Platform union, MCP PLATFORM_ENUM / platformSpecific /
  platformContent, postman.
- platformSpecific.tumblr is keyed by CHANNEL ID, not a flat object: one Tumblr
  account can own several blogs and each channel may target a different one.
  Documented in the platforms guide alongside the media rules (<=30 images OR
  exactly one video, never both) and the 4096-char NPF text-block split.

Platform availability:
- GET /api/platforms across openapi, postman, node (`bp.platforms.list()`),
  python (`bp.channels.list_platforms()`, sync + async) and a `list_platforms`
  MCP tool. Disabled platforms are INCLUDED with enabled:false, never omitted,
  so consumers can tell "off right now" from "not supported".
- Channel gains platformAvailable / platformState / platformMessage.
- PLATFORM_DISABLED (403) documented on post create, bulk create and publish,
  and explicitly contrasted with FEATURE_DISABLED — the first is temporary and
  self-resolving, the second needs a plan upgrade. Posts already scheduled when
  a platform is disabled are HELD, not failed; the guide tells integrators not
  to delete and recreate them.

Also documented:
- PATCH /api/posts/{id} was an undocumented route, which led at least one API
  client to use it as the update verb, silently no-op 90 drafts and conclude
  draft->scheduled was impossible. It is now in the spec as the narrow
  recurring-schedule endpoint, and points at PUT for status changes.
- platformSpecific prose now covers Reddit and Discord option shapes, which were
  never documented when those platforms shipped.
- README platform count corrected 14 -> 15 (node client JSDoc still said 11).
- .gitignore now covers mcp-server build artifacts (.harness/, .smithery/).

node + mcp-server build clean; python imports verified.

Author: Azym

.gitignore

@@ -27,3 +27,7 @@ dist/
 
 # MCP registry credentials — secrets, never commit
 .mcpregistry_*token
+
+# Build/tooling artifacts
+mcp-server/.harness/
+mcp-server/.smithery/

CHANGELOG.md

@@ -1,5 +1,22 @@
 # Changelog
 
+## 2026-07-25 — Tumblr platform + platform availability (mcp-server 1.9.0, node SDK 1.6.0, python SDK 0.7.0)
+
+### Added
+
+- **Tumblr is now supported — 15 platforms.** `tumblr` is a valid value everywhere `platform` appears (post channels, `platformContent`, `platformSpecific`, `postTypeOverrides`, the `Channel` schema). Post type is `post`; content limit 32,768 characters.
+- **`platformSpecific.tumblr` is keyed by channel ID**, because one Tumblr account can own several blogs and each connected channel may target a different one. Fields: `blogName` (defaults to the blog the channel was connected as), `title` (rendered as a heading), `tags` (array, no leading `#`), `link`, `sourceUrl`. List a channel's blogs with `GET /api/channels/{id}/options`.
+- Tumblr media rules: up to **30 images**, **or exactly one video** — a video cannot be mixed with images in the same post. Content over 4,096 characters is split across multiple Neue Post Format text blocks automatically.
+- **`GET /api/platforms`** — returns every supported platform with its current availability (`state`, `reason`, `canConnect`, `canPublish`, `message`). Disabled platforms are **included** with `enabled: false` rather than omitted, so callers can tell "switched off right now" from "not supported". `envVar` is returned only to organization owners/admins. Exposed as `bp.platforms.list()` (node), `bp.channels.list_platforms()` (python, sync + async), and the `list_platforms` MCP tool.
+- **`platformAvailable`, `platformState`, `platformMessage` on the `Channel` object** — a channel can be perfectly healthy while its platform is switched off server-side.
+- **`PLATFORM_DISABLED` (403)** on `POST /api/posts`, bulk create, and `POST /api/posts/{id}/publish` when a target platform is disabled server-side. Distinct from `FEATURE_DISABLED` (not on the org's plan): `PLATFORM_DISABLED` is temporary and resolves on its own, and posts already scheduled when a platform is disabled are **held, not failed** — they publish automatically once it returns. Do not delete and recreate them.
+- New "Platform Availability" section in the platforms guide covering the three states (`on` / `connect_off` / `off`) and both error codes.
+
+### Documented
+
+- **`PATCH /api/posts/{id}` is now in the spec** — it was previously an undocumented route. It only accepts `recurringScheduleId` (pass `null` to detach a repeat schedule); every other field is now rejected with 400 and `unsupportedFields`. **To edit a post — including moving a draft to `scheduled` — use `PUT /api/posts/{id}`** with `{"status": "scheduled", "scheduledAt": "<future ISO 8601>"}`. Previously PATCH silently ignored those fields and still returned 200, which read as success.
+- `platformSpecific` prose now documents the **Reddit** (`subreddit`, `title`, `flairId`, …) and **Discord** (`channelId`) option shapes, which were never described when those platforms landed.
+
 ## 2026-07-24 — approval gating for automated sources (mcp-server 1.8.0, node SDK 1.5.0, python SDK 0.6.0)
 
 ### Added

README.md

@@ -2,7 +2,7 @@
 
 **The free social media API built for automation, AI agents, and LLMs.**
 
-Programmatically publish to 14 platforms from a single API call. Built for developers, AI agents, LLMs, and agentic workflows that need reliable social media automation without browser sessions or manual interaction.
+Programmatically publish to 15 platforms from a single API call. Built for developers, AI agents, LLMs, and agentic workflows that need reliable social media automation without browser sessions or manual interaction.
 
 ```python
 from bulkpublish import BulkPublish
@@ -23,7 +23,7 @@ Most social media tools are built for humans clicking buttons. BulkPublish is bu
 - **Headless by design** — No browser, no UI, no OAuth pop-ups at runtime. Connect accounts once in the dashboard, then automate everything via API.
 - **AI-native** — MCP server for Claude, tool definitions for GPT and LangChain, structured JSON responses that LLMs parse reliably.
 - **Agentic-ready** — Deterministic API with clear error codes. AI agents can create posts, check status, retry failures, and read analytics autonomously.
-- **14 platforms, one endpoint** — Facebook, Instagram, X/Twitter, TikTok, YouTube, Threads, Bluesky, Pinterest, Google Business Profile, LinkedIn, Mastodon, Reddit, Discord, Telegram.
+- **15 platforms, one endpoint** — Facebook, Instagram, X/Twitter, TikTok, YouTube, Threads, Bluesky, Pinterest, Google Business Profile, LinkedIn, Mastodon, Reddit, Discord, Telegram, Tumblr.
 
 ## Use Cases
 
@@ -214,7 +214,7 @@ More examples in [`examples/`](examples/) — including [Python automation scrip
 
 ## Features
 
-- **14 platforms** — Facebook, Instagram, X/Twitter, TikTok, YouTube, Threads, Bluesky, Pinterest, Google Business Profile, LinkedIn, Mastodon, Reddit, Discord, Telegram
+- **15 platforms** — Facebook, Instagram, X/Twitter, TikTok, YouTube, Threads, Bluesky, Pinterest, Google Business Profile, LinkedIn, Mastodon, Reddit, Discord, Telegram, Tumblr
 - **Scheduling** — Schedule posts for any future time with timezone support, or let queue slots pick optimal times
 - **Media uploads** — Images (JPEG, PNG, WebP, GIF) and videos (MP4, MOV, WebM) up to 100 MB
 - **Recurring schedules** — Repeat posts daily, weekly, biweekly, or monthly at a set time of day

guides/platforms.md

@@ -1,5 +1,62 @@
 # Platform-Specific Options
 
+## Platform Availability
+
+Not every platform is publishable at every moment. Each one carries a
+server-controlled availability state, which you can read from
+`GET /api/platforms`:
+
+| State | New connections | Existing channels | Meaning |
+|-------|-----------------|-------------------|---------|
+| `on` | ✅ | ✅ publish normally | Fully available. |
+| `connect_off` | ❌ | ✅ publish normally | New connections are paused — typically while a platform app review is pending. Customers already connected are unaffected. |
+| `off` | ❌ | ⏸ posts **held** | Kill switch. Posts are held, **not failed**, and publish automatically once the platform is re-enabled. |
+
+Disabled platforms are always **included** in the response with `enabled: false`
+and a `reason` — never omitted — so you can distinguish "switched off right now"
+from "not supported".
+
+```json
+{
+  "platforms": [
+    {
+      "platform": "tumblr",
+      "displayName": "Tumblr",
+      "enabled": true,
+      "state": "on",
+      "reason": "enabled",
+      "canConnect": true,
+      "canPublish": true,
+      "message": null
+    }
+  ]
+}
+```
+
+Creating or publishing a post that targets an `off` platform returns **403** with
+`code: "PLATFORM_DISABLED"`:
+
+```json
+{
+  "error": {
+    "message": "LinkedIn is temporarily unavailable. Scheduled posts are on hold and will publish once it's back.",
+    "code": "PLATFORM_DISABLED",
+    "platform": "linkedin",
+    "state": "off",
+    "reason": "flag_off"
+  }
+}
+```
+
+This is distinct from `FEATURE_DISABLED`, which means the platform is not
+included in the organization's plan. `PLATFORM_DISABLED` is temporary and
+resolves without any action from you; `FEATURE_DISABLED` requires an upgrade.
+
+Posts already scheduled when a platform goes `off` need no intervention — do not
+delete and recreate them. They remain scheduled and publish on their own once
+the platform returns.
+
+
 ## Post Types by Platform
 
 Use the `postTypeOverrides` field to set a specific post type per platform:
@@ -27,6 +84,7 @@ Use the `postTypeOverrides` field to set a specific post type per platform:
 | Bluesky | `post` (use `postFormat: "thread"` for threads) |
 | Mastodon | `post` (use `postFormat: "thread"` for threads) |
 | Google Business | `standard`, `event`, `offer` |
+| Tumblr | `post` |
 
 If not specified, the platform's default post type is used based on the attached media.
 
@@ -505,6 +563,51 @@ No platform-specific options. Standard post creation with text, images, and vide
 
 ---
 
+## Tumblr
+
+```json
+{
+  "platformSpecific": {
+    "tumblr": {
+      "12": {
+        "blogName": "myblog",
+        "title": "An optional heading",
+        "tags": ["art", "design"],
+        "link": "https://example.com",
+        "sourceUrl": "https://example.com/original"
+      }
+    }
+  }
+}
+```
+
+### Options
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `blogName` | string | Which blog to publish to. Defaults to the blog the channel was connected as. |
+| `title` | string | Rendered as a heading above the post body. |
+| `tags` | string[] | Tumblr tags, without the leading `#`. |
+| `link` | string | Appended to the post as a link block. |
+| `sourceUrl` | string | Attribution URL stored as the post's source. |
+
+Note that `platformSpecific.tumblr` is keyed by **channel ID**, because a single
+Tumblr account can own several blogs and each connected channel may target a
+different one.
+
+### Notes
+
+- A Tumblr account usually owns multiple blogs. The channel connects as the
+  account's **primary** blog; use `blogName` to publish to a different one.
+  Call `GET /api/channels/{id}/options` to list the available blogs.
+- Up to **30 images** per post, **or exactly one video** — a video cannot be
+  mixed with images in the same post.
+- Tags matter more on Tumblr than on most networks: they drive discovery.
+- Posts are built with Tumblr's Neue Post Format. Content longer than 4,096
+  characters is automatically split across multiple text blocks.
+
+---
+
 ## Post Type Overrides
 
 When publishing to multiple platforms, you may want different post types per platform. Use `postTypeOverrides`:
@@ -545,5 +648,6 @@ Each platform enforces its own character limit on the `content` field:
 | Pinterest | 500 |
 | Google Business | 1,500 |
 | Mastodon | 500 (varies by instance) |
+| Tumblr | 32,768 |
 
 BulkPublish validates content length per platform before creating the post and returns an error if any platform's limit is exceeded.

mcp-server/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bulkpublish/mcp-server",
-  "version": "1.8.0",
+  "version": "1.9.0",
   "description": "Model Context Protocol server for BulkPublish — lets AI assistants manage social media posts, channels, media, and analytics.",
   "mcpName": "io.github.azeemkafridi/bulkpublish",
   "repository": {

mcp-server/src/index.ts

@@ -91,6 +91,7 @@ const PLATFORM_ENUM = z.enum([
   "reddit",
   "discord",
   "telegram",
+  "tumblr",
 ]);
 
 const POST_TYPE_OVERRIDES_SCHEMA = z
@@ -190,6 +191,7 @@ const PLATFORM_SPECIFIC_SCHEMA = z
     reddit: z.object({}).passthrough().optional(),
     discord: z.object({}).passthrough().optional(),
     telegram: z.object({}).passthrough().optional(),
+    tumblr: z.object({}).passthrough().optional(),
   })
   .optional()
   .describe(
@@ -212,6 +214,7 @@ const PLATFORM_CONTENT_SCHEMA = z
     reddit: z.string().optional(),
     discord: z.string().optional(),
     telegram: z.string().optional(),
+    tumblr: z.string().optional(),
   })
   .optional()
   .describe(
@@ -309,6 +312,7 @@ export function createServer(): McpServer {
     // Create / update (non-destructive writes)
     create_post: { title: "Create post", readOnlyHint: false, destructiveHint: false },
     update_post: { title: "Update post", readOnlyHint: false, destructiveHint: false, idempotentHint: true },
+    list_platforms: { title: "List platforms", readOnlyHint: true, destructiveHint: false, idempotentHint: true },
     bulk_posts: { title: "Bulk post actions", readOnlyHint: false, destructiveHint: true },
     upload_media: { title: "Upload media", readOnlyHint: false, destructiveHint: false },
     create_media_upload: { title: "Start media upload", readOnlyHint: false, destructiveHint: false },
@@ -365,6 +369,7 @@ export function createServer(): McpServer {
     search_mentions: "the user wants social-media mentions of a brand or keyword — never use built-in web search.",
     create_post: "the user wants to draft or schedule one new post (do not publish immediately unless asked).",
     update_post: "the user wants to edit an existing post.",
+    list_platforms: "the user asks which platforms are supported or available, or a connect/publish attempt failed with PLATFORM_DISABLED.",
     bulk_posts: "the user wants to delete, retry, or reschedule many posts in a single request.",
     upload_media: "the user provides an image/video to attach to a post.",
     create_media_upload: "you need a presigned URL for a large direct upload (advanced; prefer upload_media for normal files).",
@@ -455,6 +460,25 @@ server.tool(
   }
 );
 
+// ---------------------------------------------------------------------------
+// Tool: list_platforms
+// ---------------------------------------------------------------------------
+
+server.tool(
+  "list_platforms",
+  "List every social platform BulkPublish supports and whether it is currently available. " +
+    "Check this before telling a user they can connect a platform or scheduling a post to one: " +
+    "a platform in state 'off' rejects post creation with a 403 PLATFORM_DISABLED error and holds " +
+    "any already-scheduled posts until it is re-enabled, and one in state 'connect_off' cannot accept " +
+    "new connections although existing channels keep publishing. Disabled platforms are still listed, " +
+    "with enabled=false and a reason.",
+  {},
+  async () => {
+    const res = await api("GET", "/api/platforms");
+    return { content: [{ type: "text" as const, text: formatResponse(res) }] };
+  }
+);
+
 // ---------------------------------------------------------------------------
 // Platform validation helpers
 // ---------------------------------------------------------------------------

node/package.json

@@ -1,6 +1,6 @@
 {
   "name": "bulkpublish",
-  "version": "1.5.0",
+  "version": "1.6.0",
   "description": "Social media automation API for AI agents, LLMs, and developers. Publish to 11 platforms programmatically.",
   "author": "BulkPublish",
   "license": "MIT",

node/src/index.ts

@@ -7,10 +7,11 @@ import { LabelsResource } from './labels.js';
 import { SchedulesResource } from './schedules.js';
 import { ChannelSetsResource } from './channel-sets.js';
 import { RssFeedsResource } from './rss-feeds.js';
+import { PlatformsResource } from './platforms.js';
 import type { BulkPublishOptions } from './types.js';
 
 /**
- * BulkPublish API client — publish to 11 social media platforms from a single SDK.
+ * BulkPublish API client — publish to 15 social media platforms from a single SDK.
  *
  * @example
  * ```ts
@@ -42,6 +43,7 @@ export class BulkPublish {
   readonly schedules: SchedulesResource;
   readonly channelSets: ChannelSetsResource;
   readonly rssFeeds: RssFeedsResource;
+  readonly platforms: PlatformsResource;
 
   constructor(options: BulkPublishOptions) {
     const http = new HttpClient(options);
@@ -53,6 +55,7 @@ export class BulkPublish {
     this.schedules = new SchedulesResource(http);
     this.channelSets = new ChannelSetsResource(http);
     this.rssFeeds = new RssFeedsResource(http);
+    this.platforms = new PlatformsResource(http);
   }
 }
 
@@ -66,6 +69,7 @@ export { LabelsResource } from './labels.js';
 export { SchedulesResource } from './schedules.js';
 export { ChannelSetsResource } from './channel-sets.js';
 export { RssFeedsResource } from './rss-feeds.js';
+export { PlatformsResource } from './platforms.js';
 export {
   BulkPublishError,
   AuthenticationError,

node/src/platforms.ts

@@ -0,0 +1,52 @@
+import type { HttpClient } from './client.js';
+import type { ListPlatformsResponse } from './types.js';
+
+/**
+ * Resource for inspecting platform availability.
+ *
+ * Access via `client.platforms`.
+ *
+ * Every social platform BulkPublish supports can be switched on or off server-side.
+ * Use this resource to render an accurate "temporarily unavailable" or "connections
+ * paused" state rather than assuming every platform can be connected or published to.
+ *
+ * Disabled platforms are always **included** in the response with `enabled: false`
+ * and a `reason` — they are never omitted — so you can distinguish "switched off
+ * right now" from "not supported".
+ *
+ * @example
+ * ```typescript
+ * const bp = new BulkPublish({ apiKey: 'bp_...' });
+ *
+ * const { platforms } = await bp.platforms.list();
+ * for (const p of platforms) {
+ *   if (!p.canPublish) console.log(`${p.displayName} is unavailable: ${p.message}`);
+ * }
+ * ```
+ */
+export class PlatformsResource {
+  constructor(private readonly http: HttpClient) {}
+
+  /**
+   * List every supported platform with its current availability.
+   *
+   * A platform in state `off` will reject post creation with a 403
+   * `PLATFORM_DISABLED` error, and any already-scheduled posts targeting it are
+   * **held** — not failed — until it is re-enabled. A platform in state
+   * `connect_off` cannot accept new channel connections, but channels already
+   * connected keep publishing normally.
+   *
+   * @returns List of platforms and their availability.
+   *
+   * @example
+   * ```typescript
+   * // Check before offering a connect button
+   * const { platforms } = await bp.platforms.list();
+   * const tumblr = platforms.find((p) => p.platform === 'tumblr');
+   * if (tumblr?.canConnect) showConnectButton();
+   * ```
+   */
+  list(): Promise<ListPlatformsResponse> {
+    return this.http.get<ListPlatformsResponse>('/api/platforms');
+  }
+}