Sprout Social MCP Server

A Model Context Protocol server for Sprout Social analytics, inbox messages, social listening, content publishing, media uploads, and support cases. It works with Claude, GPT, and other MCP clients.

Why this exists

Sprout Social's API can answer questions and create drafts without a trip through the dashboard for each operation. This server makes those calls available to MCP clients and adds a few workflows for common reporting, campaign, and case-triage jobs.

Related MCP server: @posteverywhere/mcp

What you can do

Ask your AI assistant to work with Sprout Social directly:

• "How did our Instagram perform last month?" Pull profile impressions, engagements, and follower growth for the date range.

• "Show me our top posts on LinkedIn this quarter." Query post analytics and sort by the metrics you choose.

• "What are people saying about our brand?" Search listening topics by sentiment, network, language, or text.

• "Draft a post announcing our product launch across all channels." Create draft posts with media, scheduling, and tags.

• "Are there any open support cases assigned to me?" Filter cases by status, priority, assignee, and queue.

28 tools across 6 API domains

Authentication: 1 tool

Metadata: 8 tools

Discover and enumerate your Sprout Social account structure.

Analytics: 4 tools

Query performance metrics across profiles and individual posts. Supports all major networks: Instagram, Facebook, X (Twitter), LinkedIn, YouTube, Pinterest, TikTok, Threads, and Bluesky.

Messages: 2 tools

Listening: 3 tools

Publishing: 8 tools

Create draft content and manage media uploads of any size.

Cases: 2 tools

Quick start

Prerequisites

• Node.js 20.19+

• A Sprout Social account with API access enabled (request access)

Install from npm

npm install -g @oliverames/sprout-mcp-server

Or run directly with npx:

npx @oliverames/sprout-mcp-server

Install from source

git clone https://github.com/oliverames/sprout-mcp-server.git
cd sprout-mcp-server
npm ci && npm run build

Authentication

There are two ways to connect. Pick the one that fits how the server runs.

Option A: token or machine auth

Use this when the server runs headless (CI, a scheduled job, a shared host).

A1. API token (simplest). Generate a token in Sprout under Settings → Global Features → API Page → API Token Management.

export SPROUT_API_TOKEN=your-token-here

A2. OAuth 2.0 machine-to-machine. Client-credentials grant; the server acquires and refreshes the token automatically.

export SPROUT_CLIENT_ID=your-client-id
export SPROUT_CLIENT_SECRET=your-client-secret
export SPROUT_ORG_ID=your-org-id

Option B: sign in with Sprout

Use this to connect your own Sprout account. You sign in at Sprout's real login page in the browser, so the server never sees your password and your normal 2FA/SSO applies. The interactive flow uses PKCE, so it needs only a client ID, with no client secret to configure or store.

1. Set your OAuth client ID (generate it under Settings → Global Features → API → OAuth Client Management):

   export SPROUT_CLIENT_ID=your-client-id

2. Run the login command and sign in when the browser opens:

   npm run login

The CLI starts a local callback server, opens your browser, and on success saves your session to ~/.sprout-mcp-auth.json. After that, run the server with no further configuration:

npm start

The server loads your saved session and refreshes tokens automatically in the background.

1Password integration

If credentials are not set in the environment, the server automatically attempts to resolve them from 1Password CLI:

This means you can skip setting env vars entirely if you have op installed and a service account or session active. The fallback adds ~1-2s to startup and is silently skipped if 1Password is unavailable.

Configuration

Claude Desktop and Claude Code

Add to your MCP settings (e.g., .claude/settings.json):

{
  "mcpServers": {
    "sprout": {
      "command": "npx",
      "args": ["@oliverames/sprout-mcp-server"],
      "env": {
        "SPROUT_API_TOKEN": "your-token-here"
      }
    }
  }
}

From a local clone

{
  "mcpServers": {
    "sprout": {
      "command": "node",
      "args": ["/path/to/sprout-mcp-server/dist/index.js"],
      "env": {
        "SPROUT_API_TOKEN": "your-token-here"
      }
    }
  }
}

Any MCP client

The server communicates over stdio transport. Point any MCP-compatible client at node dist/index.js (or npx @oliverames/sprout-mcp-server) with the appropriate environment variables.

Multi-customer support

On startup, the server auto-discovers your customer ID from the API. If your token has access to multiple customers, specify which one:

export SPROUT_CUSTOMER_ID=123456

Every tool also accepts an optional customer_id parameter to override the default for a request, which is useful for agencies managing multiple brands.

Response formats

All tools accept a response_format parameter:

Supported networks

Analytics, messages, and publishing support all networks connected in your Sprout Social account:

Built-in reliability

The server handles the operational details so you don't have to:

• A sliding-window throttle stays under Sprout's 60 request-per-minute limit, with a soft cap at 55.

• Requests retry with exponential backoff after 429, 500, 503, and 504 responses.

• Media uploads and multipart completions poll automatically after a 202 response.

• Zod validates dates, profile limits, and required fields before a request reaches Sprout.

• The server starts without credentials and keeps its tool catalog visible with setup instructions.

• Large responses are truncated before they overwhelm the client's context window.

Tool coverage

The server maps 20 Sprout API endpoints to direct tools and adds seven workflow tools. The authentication status tool brings the protocol-visible catalog to 28. An integration test lists the tools through MCP and checks every name.

Known limitations

These are Sprout Social API constraints, not server limitations:

Development

npm ci               # Install locked dependencies
npm run build        # Compile TypeScript
npm test             # Run the test suite
npm run check        # Build, test, audit, and inspect the npm package
npm run test:watch   # Watch mode
npm run dev          # TypeScript watch mode
npm start            # Start the server

See CONTRIBUTING.md for pull request guidance, SECURITY.md for private vulnerability reporting, and RELEASING.md for the release checklist.

Project structure

src/
├── index.ts              # Entry point: auth, customer discovery, tool registration
├── constants.ts          # API URLs, rate limits, validation constraints
├── types.ts              # Shared TypeScript interfaces
├── schemas/
│   └── common.ts         # Reusable Zod schemas for tool input validation
├── services/
│   ├── auth.ts           # API token + OAuth M2M token management
│   ├── api-client.ts     # HTTP client with retries, rate limiting, 202 polling
│   ├── filter-builder.ts # Translates typed params → Sprout's filter DSL (eq, neq, in, gt, match, exists)
│   └── formatter.ts      # Markdown/JSON formatting + truncation
└── tools/
    ├── catalog.ts        # Catalog inventory and registration
    ├── metadata.ts       # 8 account structure tools
    ├── analytics.ts      # 4 profile, post, and reporting tools
    ├── messages.ts       # 2 inbox message tools
    ├── listening.ts      # 3 topic and trend tools
    ├── publishing.ts     # 8 draft and media tools
    └── cases.ts          # 2 case management tools

Architecture

All tool handlers are pure functions with the signature:

handler(client: ApiClient, customerId: number, params: T) → Promise<ToolResponse>

This makes every handler independently testable with a mock ApiClient. The catalog test uses an in-memory MCP transport to check the list clients receive.

The filter builder translates friendly parameters into Sprout's custom filter DSL (field.op(values)), keeping the DSL syntax internal to the server.

License

MIT

Sprout Social is a trademark of Sprout Social, Inc. This independent project is not affiliated with or endorsed by Sprout Social, Inc.