Your AI Agent. Your Social Media Manager.

A standardized tool layer

Model Context Protocol (MCP) is an open specification from Anthropic that defines how language models discover and call external tools. Instead of every app shipping its own custom “function calling” layer, MCP gives every model the same wire format: a JSON-RPC handshake, a tool registry, and structured input/output schemas.

Beyond REST APIs

A traditional REST API requires a developer to write client code, parse OpenAPI specs, and wire up auth. With MCP, the AI agent reads the tool catalog at runtime, understands argument shapes from the schema, and decides on its own which tool to call — no glue code, no custom adapters, no per-integration prompt engineering.

Not a chatbot, not a plugin

Chatbots reply with text. Plugins live inside one vendor's walled garden. MCP servers are local processes you run yourself, exposed to any compatible client. CodivUpload's MCP server runs on your machine, holds your API key, and gives whichever model you're using the same set of tools — today Claude, tomorrow the next state-of-the-art agent.

Anthropic's reference MCP implementation. Works on Claude Desktop 0.7+ (macOS, Windows, Linux) and Claude Code CLI 1.0+. Config lives in ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or the equivalent path on other OSes. Claude Code reads .mcp.json from the project root, which makes per-project API keys easy.

claude mcp add codivupload \
  -e CODIVUPLOAD_API_KEY=cdv_... \
  -- npx -y codivupload-mcp

Cursor 0.45+ ships first-class MCP support. Configure servers under Settings → MCP, or drop a .cursor/mcp.json in your repo. The IDE will auto-restart the server on config changes — useful when rotating API keys.

{
  "mcpServers": {
    "codivupload": {
      "command": "npx",
      "args": ["-y", "codivupload-mcp"],
      "env": { "CODIVUPLOAD_API_KEY": "cdv_..." }
    }
  }
}

Cline is the most popular open MCP client for VS Code. Add CodivUpload through the MCP Servers tab in the Cline sidebar — same JSON shape as Cursor and Claude Desktop. Works with Anthropic, OpenAI, and any model Cline supports as a backend.

# In Cline Settings -> MCP Servers
{
  "command": "npx",
  "args": ["-y", "codivupload-mcp"],
  "env": { "CODIVUPLOAD_API_KEY": "cdv_..." }
}

Continue.dev, Zed, Windsurf, Sourcegraph Cody, and a growing list of agents all read the same MCP spec. If your client supports stdio MCP servers and lets you set environment variables, CodivUpload works — no special integration required. Run the server manually with npx codivupload-mcp and connect over stdio.

CODIVUPLOAD_API_KEY=cdv_... \
  npx -y codivupload-mcp

Bulk operations

Hand the agent a CSV of 200 posts with platform tags, captions, media URLs, and dates. It loops through, validates each row, dispatches schedule_post calls, and reports back with successes, failures, and quota remaining. No bulk-import UI required — the agent is the bulk-import UI.

Cross-platform analytics queries

“Which platform gave me the highest engagement rate last month, and what did the top three posts have in common?” The agent calls get_posts with a date filter, aggregates across platforms, runs the qualitative analysis itself, and gives you a one-paragraph answer instead of a dashboard you have to read.

Workspace management

Audit token health, find disconnected platforms, list scheduled posts that haven't fired, and pull a quota usage report — all through natural language, all without opening the dashboard. Useful for agencies running 20+ client workspaces who want a daily standup summary delivered into their chat client.

Team coordination workflows

Pair the MCP server with another MCP server (Linear, Notion, Slack) and the agent can move work across systems: pull the next approved caption from Notion, schedule it on the right channels, post a confirmation to Slack, and update the Linear ticket to “Shipped.” Multi-tool orchestration without writing a Zapier flow.