About aidelly
Aidelly is a Claude Code plugin for agencies managing social media across connected platforms. It uses MCP tools for publishing, approvals, media uploads, and engagement analytics, with OAuth sign-in.
Key features
- 01Create and schedule posts
- 02Manage post approval workflows
- 03Review engagement analytics
- 04Upload images and videos
- 05Support 11 social platforms
Use cases
- 01Schedule posts across social accounts
- 02Review pending posts before publishing
- 03Compare engagement and top-performing content
MCP support
Connects Claude Code to Aidelly’s remote MCP server over streamable HTTP. OAuth 2.1 with PKCE handles authentication; tools cover posts, approvals listing, analytics, and media uploads.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://app.aidelly.ai/api/mcp/public-apiclaude mcp add --transport http wilzer-aidelly https://app.aidelly.ai/api/mcp/public-api{"mcpServers":{"wilzer-aidelly":{"type":"http","url":"https://app.aidelly.ai/api/mcp/public-api"}}}Open in Cursor{"servers":{"wilzer-aidelly":{"type":"http","url":"https://app.aidelly.ai/api/mcp/public-api"}}}Open in VS Code{"mcpServers":{"wilzer-aidelly":{"type":"http","url":"https://app.aidelly.ai/api/mcp/public-api"}}}{"mcpServers":{"wilzer-aidelly":{"serverUrl":"https://app.aidelly.ai/api/mcp/public-api"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
150| Parameter | Type | Required | Description |
|---|---|---|---|
| body | object | No | Optional body override for endpoints with sparse parameter schemas. |
| tags | array | No | |
| text | string | No | Document text. Supply exactly one of text or source_url. |
| query | object | No | Optional query overrides for endpoints with sparse parameter schemas. |
| title | string | No | |
| brand_id | string | No | |
| source_url | string | No | Public http(s) URL of a PDF, DOCX, TXT or Markdown file (10 MB max). Web pages (HTML) are not accepted; send their text as text instead. |
| description | string | No | |
| workspace_id | string | No | Workspace to operate in: the `id` of a workspace returned by aidelly_list_workspaces. Optional for read operations; required when creating content. |
| document_type | enum | No | brand_guidelines | logo | template | presentation | report | contract | media_kit | style_guide | content_calendar | other |
| idempotency_key | string | Yes |
Details
From the project READMEAidelly Claude Code Plugin
Automate Aidelly social media management directly from Claude Code. Create, schedule, and manage posts across all connected platforms with AI-powered MCP tools.
Features
- Create & Schedule Posts – Publish instantly or schedule for later across multiple platforms
- Approval Workflows – Require and manage post approvals before publishing
- Analytics – Pull engagement metrics and top-performing content
- Multi-Platform Support – Instagram, Facebook, LinkedIn, Twitter/X, TikTok, YouTube, Pinterest, Bluesky, Threads, Mastodon, Google Business
- Media Upload – Fetch and upload images/videos directly
Installation
Sign-in is one click. The Aidelly MCP server supports OAuth 2.1 with PKCE and dynamic client registration. Claude walks you through an Aidelly login and consent screen, so there's no key or workspace ID to copy, and the plugin never reads credentials from your machine.
Install via Claude Code Plugin Marketplace
/plugin marketplace add Aidelly/claude-plugin
Then sign in to Aidelly (see Setup below).
Manual Installation (Development)
Clone this repository and point Claude Code to the local plugin path:
git clone https://github.com/Aidelly/claude-plugin
cd claude-plugin
# Add to Claude Code plugin settings
Note: Ensure
.mcp.jsonis committed in your local clone. The Aidelly monorepo ignores*.mcp.jsonfiles, so the publicAidelly/claude-pluginrepository explicitly commits it. If you see a missing.mcp.jsonafter cloning, rungit check-ignore .mcp.json— if no output, it's been properly committed.
Setup
1. Sign in
The first time Claude uses an Aidelly tool, Claude Code asks you to sign in:
- Run
/mcpin Claude Code and choose aidelly → Authenticate (or start with any Aidelly request). - Your browser opens the Aidelly login and consent screen. Log in and approve access.
- Return to Claude Code. The server shows as connected.
Signing in needs an Aidelly plan that includes API access. The token is issued to your account and expires; sign in again from /mcp when it does.
2. Pick a workspace
If you manage several workspaces, ask Claude to list them (aidelly_list_workspaces) and say which one to use. Claude passes the workspace to each tool, so there's nothing to configure.
3. Verify the connection
In Claude Code, run:
List my Aidelly workspaces and connected accounts.
Claude should return your workspace data.
Usage
Create a Post
Create a new post on Instagram with the text "Hello world!" and upload media.
Schedule a Post
Schedule a post for tomorrow at 2 PM with approval required.
Check Pending Approvals
List all posts waiting for approval and approve the first one.
Pull Analytics
Get the top 10 posts from the last 30 days and their engagement metrics.
Skill Documentation
Full endpoint reference and examples: skills/aidelly-social/SKILL.md
Configuration
.mcp.json
The plugin wires to Aidelly's remote MCP server over streamable HTTP. There are no headers and no environment variables; OAuth handles sign-in:
{
"mcpServers": {
"aidelly": {
"type": "http",
"url": "https://app.aidelly.ai/api/mcp/public-api"
}
}
}
plugin.json
Metadata for the Claude Code plugin directory:
{
"name": "aidelly",
"description": "Aidelly social content automation for Claude Code — create posts, schedule content, check approvals, and pull analytics.",
"author": { "name": "Aidelly", "url": "https://aidelly.ai" },
"homepage": "https://aidelly.ai",
"repository": "https://github.com/Aidelly/claude-plugin",
"privacyPolicyUrl": "https://www.aidelly.ai/privacy-policy",
"icon": "./assets/icon.png",
"version": "0.2.2"
}
Troubleshooting
"Unauthorized" Error
- Run
/mcp, choose aidelly → Authenticate, and sign in again (the token may have expired) - Confirm your Aidelly plan includes API access
"Forbidden" Error
- Make sure the workspace Claude is using is one your Aidelly account can access
- Ask Claude to list your workspaces and pick the right one
Rate Limit Exceeded
- The server allows 120 tool calls per minute per signed-in user
- Retry after a few seconds with exponential backoff
Platform-Specific Issues
- Instagram/TikTok deletion: These platforms don't allow post deletion via API
- LinkedIn org routing: Double-check that the author target matches your account type (member vs. organization)
- Text too long: Verify text length against platform limits in the skill documentation
Architecture
.claude-plugin/
└─ plugin.json # Plugin metadata
.mcp.json # MCP server configuration
assets/
└─ icon.png # 512×512 listing icon
skills/
└─ aidelly-social/
└─ SKILL.md # Tool documentation & examples
API Endpoints
The plugin connects to https://app.aidelly.ai/api/mcp/public-api (Aidelly's MCP HTTP endpoint).
Authentication: OAuth 2.1 (PKCE + dynamic client registration), discovered from the server's .well-known/oauth-protected-resource metadata. Tool calls without a token return 401 with a WWW-Authenticate challenge, which starts the sign-in.
Tools exposed by the server:
aidelly_create_post– Instant post publishaidelly_create_scheduled_post– Schedule a postaidelly_list_pending_approvals– Check pending approvals- Approving or rejecting posts happens in the Aidelly app; the hosted server does not expose approval actions
aidelly_get_analytics_summary– Pull engagement metricsaidelly_get_post– Get post detailsaidelly_upload_media– Upload image/video- Full list of available tools
Publishing & Distribution
This repository is designed to be published to Claude Code plugin marketplaces. Follow the checklist below for each distribution channel.
Distribution Channels
1. Anthropic Claude Code Plugin Directory
Status: [Pending / In progress / Published]
Submission checklist:
- README.md complete with features, installation, setup
- .claude-plugin/plugin.json valid and descriptive
- .mcp.json correctly configured for production MCP endpoint
- Skills documentation comprehensive with examples
- All JSON files validate (no parse errors)
- Version bumped to match release (semver)
- LICENSE file included (if required)
- Documentation reviewed for accuracy
How to submit:
- Contact Anthropic at [email protected] with:
- GitHub repository link
- Brief description of plugin capabilities
- API authentication requirements
- Follow Anthropic's submission process
- Monitor distribution portal for approval status
2. Official MCP Registry (Smithery)
Status: [Pending / In progress / Published]
Registry: https://smithery.ai or https://registry.smithery.ai
Submission checklist:
- Repository public on GitHub
- .mcp.json follows MCP spec v1.0+
- Version in package.json matches .claude-plugin/plugin.json
- README includes authentication setup
- Endpoint URL is production-ready (https://)
- Rate limits documented
- Error handling documented
- Test credentials available for verification (optional)
How to submit:
- Fork the MCP registry (https://github.com/smithery/smithery-registry)
- Add entry in registries.json (or equivalent)
- Include metadata: name, version, description, auth requirements, URL
- Open pull request
- Address review feedback
3. Listed on
Glama – https://glama.ai/mcp/servers/Aidelly/claude-plugin
The badge at the top of this README reflects Glama's live quality score for this server.
4. Alternative MCP Registries
mcp.so – https://mcp.so PulseMCP – https://pulsemcp.com
Submission checklist (each):
- Plugin discoverable via their search
- Correct metadata (name, description, author)
- Production MCP endpoint working
- Authentication method documented
- Contact/support information provided
How to submit:
- Each registry has its own submission process (check their websites)
- Generally requires GitHub repo + metadata + contact info
Pre-Release Checklist
Before submitting to any marketplace:
# 1. Validate JSON files
jq . .claude-plugin/plugin.json
jq . .mcp.json
# 2. Verify the MCP endpoint answers (no token needed for initialize)
curl -s -X POST https://app.aidelly.ai/api/mcp/public-api \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","method":"initialize","id":1}'
# 3. Verify the OAuth challenge on a tool call (expect HTTP 401 + WWW-Authenticate)
curl -s -D - -o /dev/null -X POST https://app.aidelly.ai/api/mcp/public-api \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"aidelly_list_workspaces","arguments":{}}}'
# 4. Review skill documentation
cat skills/aidelly-social/SKILL.md | wc -l # Should be > 200 lines
Release Notes
When bumping version in .claude-plugin/plugin.json and README:
## Version 0.2.2
- Uses the renamed tool IDs (for example `aidelly_create_content_automation`); the old IDs stay callable as hidden aliases
- The approval workflow approves in the Aidelly app, because the hosted server does not expose approval actions
## Version 0.2.1
- Adds `privacyPolicyUrl` (https://www.aidelly.ai/privacy-policy) to plugin.json
## Version 0.2.0
- `.mcp.json` declares `"type": "http"` so Claude Code loads the remote server
- Sign-in is OAuth only; no credentials are read from the user's machine
- Adds a 512×512 listing icon
## Version 0.1.0
**Initial release**
- Create, schedule, and publish posts across multiple platforms
- Approval workflow integration
- Analytics summary pull
- Media upload support
- 11 platforms supported
Support
- Issues: https://github.com/Aidelly/claude-plugin/issues
- Docs: https://docs.aidelly.ai
- Email: [email protected]
License
[License TBD – check Aidelly open-source policy]
Maintainers
- Aidelly Team (@aidelly)
Last updated: 2026-10-01
Plugin version: 0.2.2
MCP protocol version: 2024-11-05+
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.