Introduction¶
When performing social media operations with DeepSeek Harness (dsh), a common approach is to let the model handle only the copywriting, while tasks like selecting images, scheduling, publishing, and checking data still require switching back and forth between various platform backends. If you want the model to complete drafting, scheduling, publishing, viewing analytics, and handling Direct Messages and comments in a single chat, check out the bulkpublish/dsh-plugin-bulkpublish plugin. It is maintained by BulkPublish, connects dsh to BulkPublish’s MCP server, and allows the model to directly call 32 social media operation tools.
What is this¶
dsh-plugin-bulkpublish is a dsh plugin that connects to the BulkPublish MCP server (URL https://mcp.bulkpublish.com). After installation, the model can draft, schedule, publish, and analyze social media content, upload and browse media, and handle inbox conversations and post comments within a single dsh session. The tools appear to the model as mcp__bulkpublish__<tool>.
Its implementation loads the built-in @deepseek-ai/dsh-mcp-client as a sub-plugin using the streamable-http transport method in the plugin’s apply(ctx, config).
Core Features¶
There are 15 supported platforms in total: Facebook, Instagram, X/Twitter, TikTok, YouTube, Threads, Bluesky, Pinterest, Google Business Profile, LinkedIn, Mastodon, Discord, Telegram, Tumblr, and Snapchat.
The MCP server exposes 32 tools via tools/list, grouped by purpose as follows:
| Category | Tools |
|---|---|
| Posts | create_post, update_post, get_post, list_posts, delete_post, publish_post, retry_post, get_queue_slot |
| Approval | approve_post, reject_post |
| Channels | list_channels, get_channel_options, update_channel |
| Media | upload_media, list_media, create_media_upload, finalize_media_upload |
| Labels | list_labels |
| Analytics | get_analytics, get_post_metrics |
| Inbox | list_conversations, get_conversation, reply_to_conversation, update_conversation |
| Comments | list_post_comments, reply_to_comment, moderate_comment |
| Interactive views | compose_post, view_posts, view_channels, view_media, view_analytics |
Installation and Usage¶
First, register an account on bulkpublish.com and connect the channels you want to use. Then, go to the developer dashboard app.bulkpublish.com/developer to generate an API key, which starts with bp_.
Export the key before starting dsh:
export BULKPUBLISH_API_KEY=bp_your_key_here
The authentication method sends the key as a bearer token (Authorization: Bearer <key>), reads it from the environment variable at runtime, does not write it to the profile, and does not distribute it with the package. If the key is not set, the plugin will log a warning and not register any tools.
Install via npm:
dsh plugin --profile web add dsh-plugin-bulkpublish
Install from source:
dsh plugin --profile web add github:bulkpublish/dsh-plugin-bulkpublish
After installation, the following entry will be written to the profile (cordis.patch.yml):
- id: bulkpublish
name: 'dsh-plugin-bulkpublish'
config:
url: 'https://mcp.bulkpublish.com'
Optional configuration items: apiKeyEnv (environment variable name, default BULKPUBLISH_API_KEY), serverName (tool namespace, default bulkpublish), toolCallTimeoutMs (default 60000).
Typical Usage¶
After installation, you can directly issue commands in the dsh chat using natural language. Here are example prompts given in the README:
- List the connected channels and tell me which ones need to be reconnected.
- Draft a Spring promotion copy for LinkedIn and X, and schedule it for the next available queue slot.
- Upload
https://example.com/banner.pngand schedule an Instagram post with the image for Friday at 10:00. - Show last month’s analytics data broken down by platform.
- Retry the platforms where the post failed to publish yesterday.
- Display unread inbox conversations and draft a reply for my review.
Applicable Scenarios and Notes¶
Suitable for dsh developers and operators who need to post and view data across multiple social media platforms, especially scenarios where content creation, scheduled publishing, data analysis, and interactive replies need to be consolidated into a single AI agent session.
Note three points before use. First, publishing, replying, and comment moderation act immediately on real accounts, so the model should review these calls before approving them. Second, the plugin runs with the permissions of the current dsh process; check the source code and license before installing; this plugin uses the MIT license and originates from bulkpublish/dsh-plugin-bulkpublish. Third, this plugin is not an official component of DeepSeek/Fellow AI, but a third-party plugin in the community directory.
Conclusion¶
Overall, the value of dsh-plugin-bulkpublish lies in unifying the full social media workflow tools into a single MCP server, allowing the dsh model to cover posting, analytics, and interaction across 15 platforms with a single set of tools. The directory page is at https://www.skillhub.cn/plugins/bulkpublish/dsh-plugin-bulkpublish, and the source code is at https://github.com/bulkpublish/dsh-plugin-bulkpublish.
