Meltwater MCP
In this guide we'll introduce you to Meltwater MCP - a single remote MCP server that exposes your Meltwater assets and data as tools your AI agents can call.
Meltwater MCP is accessible to API customers if they have a Meltwater MCP package in their subscription. Tools available within the MCP server depends on the products in your Meltwater package. If you are interested in access, please talk to your account manager, or our sales team.
What is Meltwater MCP?
MCP (Model Context Protocol) has quickly become the recognised standard for connecting remote tools and data to large language models (LLMs) and the agents built on top of them.
Meltwater MCP is a single remote MCP server that exposes a set of tools for accessing:
- Your Meltwater assets - for example your saved searches, tags, and other objects you have configured in the Meltwater application.
- Meltwater data - for example news and social mentions, analytics, and insights drawn from the Meltwater platform.
Because the server speaks MCP, any MCP-compatible client or agent can discover these tools and call them on demand. You bring the reasoning and orchestration; Meltwater MCP provides the tools.
Who is it for?
Meltwater MCP is built primarily for developers building their own custom agents. If you are writing an agent - in the OpenAI Agents SDK, the Anthropic SDK, LangGraph, or your own framework - and you want it to reach Meltwater data as it reasons, Meltwater MCP is the integration layer.
For non-developer workflows, we also provide connectors for Claude and ChatGPT, so end users can add Meltwater as a tool source directly within those assistants.
When you are ready to connect, see the Connecting a custom agent to Meltwater MCP guide.
Meltwater MCP vs the Mira API
Meltwater offers two MCP integration options, and it is worth being clear about when to use each. Both are exposed as remote MCP servers, but they are designed for different jobs.
| Meltwater MCP | Mira API (MCP) | |
|---|---|---|
| Designed for | Tool access for agents you orchestrate yourself | Orchestrated outcomes produced by Meltwater's own AI |
| You provide | The reasoning, planning, and orchestration | A natural-language prompt |
| Meltwater provides | A set of granular tools over your assets and data | An AI assistant that plans and answers for you |
| Best when | You are building a custom agent and want direct, composable access to Meltwater | You want a finished answer, brief, or report without building the orchestration |
| Endpoint | https://api.meltwater.com/v2/mcp | https://api.meltwater.com/mcp |
In short:
- Reach for Meltwater MCP when you want to drive the workflow and simply need reliable tool access to Meltwater from your own agent.
- Reach for the Mira API when you want Meltwater's assistant to do the orchestration and hand you a grounded outcome.
Both products will run alongside each other, so you can choose the right approach - or combine them - for each use case.
Key concepts
Tools
A tool is a single capability the server exposes to an agent - for example finding one of your saved searches, or retrieving mentions from the Meltwater platform. Each tool ships with a description that tells the LLM what it does and when to use it, so your agent can select and call the right one.
Tool discovery
Meltwater MCP exposes its tools through the standard MCP tool-discovery mechanism, so your client lists the available tools at runtime rather than you hard-coding them. The exact set of tools available to you depends on the products in your Meltwater package.
Not sure which tools your token exposes? Use the Tools Explorer to connect with your API token and browse the MCP tools available to you - each tool's name, description, and parameters - without writing any code.
Authentication
Meltwater MCP is authenticated with your Meltwater API token today, with OAuth 2.0 planned for later this year. See Connecting a custom agent to Meltwater MCP for details.
Next steps
- API Credentials - create the API token you'll use to authenticate.
- Connecting a custom agent to Meltwater MCP - connect your custom agent to the server in code.