Understanding MCP Credits vs. Tokens
With the new MCP Usage page, Account Admins can now see exactly where their Intelligence Credits go. That raises a fair question: “My AI assistant uses tokens. Meltwater uses credits. Aren’t they the same thing?” We’re here to show you how they’re different.
TL;DR: Tokens are used in every message you send. Credits only get used when your assistant uses Meltwater tools.
| Tokens | Intelligence Credits | |
|---|---|---|
| Who bills you | Your AI provider | Meltwater |
| What uses them | Every message you send and every reply | Only when your assistant calls a Meltwater tool |
| What drives the amount | How long the conversation is | Which Meltwater tools get called |
| Where to see them | Your AI provider’s usage or billing settings | Account > MCP Usage, or ask your assistant |
Quick check:
Asked for fresh data from Meltwater? That uses tokens, and usually credits.
Asked to reformat, summarize or rewrite what you already have? That uses tokens only.
Tokens: Billed by Your AI Provider
Tokens measure how much text your AI assistant (Claude, ChatGPT, Copilot and others) reads and writes. Your AI provider bills you for them. Each time you send a message, the assistant:
- Reads your message
- Rereads the whole conversation so far
- Writes its reply
Each of those steps uses tokens, whether or not Meltwater is involved.
Intelligence Credits: Billed by Meltwater
Credits measure the Meltwater intelligence your assistant pulls in using Meltwater MCP. They’re used from your Meltwater credit balance, only when the assistant calls a Meltwater tool. A tool is one task the AI Assistant assigns to Meltwater, like building a search or pulling analytics. The AI assistant chooses the tools needed for each task, but you can guide it.
Some tools are charged by credit cost, depending on how much work the tool does. Others are free. You can recognize a tool when you expand its dropdown to see a “request” (might look different in your AI environment).
Multiple Tool Calls Add Up Credits
For the next example, you can ask a question like, “How has Poppi been performing in the last 7 days?” To answer, the assistant calls one Meltwater tool, Aggregate document metrics, to count the Poppi mentions by platform and by day. That call uses 25 credits. The assistant then builds the table and writes the summary. That part costs no credits.
Tools that cost credits use an amount assigned at the tool level. The credits are only used in prompts that call Meltwater MCP tools.
Follow-Up Messages Use Tokens, Not Always Credits
Next, you might ask your AI assistant to “Refresh this weekly brief.” It calls Meltwater for fresh data, and those tool calls use credits. Then you say “Now reformat that as bullets.” That second request never reaches Meltwater, because the assistant already has the data. It uses tokens, because the assistant has to reread the thread and rewrite the brief. It uses no credits.
Recap: Prompts Use Text, Credits Are Used When Calling Meltwater
Hope that clarifies it: Prompt length drives tokens. Meltwater calls drive credits.
That’s why a long, extensive conversation can use plenty of tokens and zero credits. It’s also why a four-word prompt can use more credits than you’d expect if it calls several heavyweight tools.
Try it yourself
Type a prompt, pick a model, and see the estimated tokens. Turn on Meltwater MCP to see the credits a tool call would use.
Open the calculator in a new window
Good to Know
Where to track credits: Account Admins can open Account > MCP Usage to see credit spend by user, by tool and over time.
Anyone on your account with Meltwater MCP can also ask their assistant for usage, even users who aren’t admins. Ask something like “What’s my credit usage?” or “How many credits did this cost?” The assistant uses the account usage tool (account_get_credit_usage) to pull the numbers. Note: Checking usage doesn’t use credits.
By default, the answer covers the current month for everyone on the account. You can ask for a different date range, a specific tool, or a breakdown by month, tool, or user. To check one conversation, ask “How many credits were used?” in that same conversation.
Tips for the account usage tool: Results are most accurate for future conversations. If you try account_get_credit_usage in older conversations, it may not be accurate to future credit measurements.
Where to track tokens: Check your AI provider’s usage or billing settings. Meltwater doesn’t see or charge for tokens.
Making credits go further: Follow-up requests like reformatting, summarizing or rewriting usually don’t call Meltwater again. Being specific up front also helps the assistant pick the right tool the first time.
Troubleshooting
The assistant used more credits than I expected. Expand the tool dropdowns in the conversation to see which Meltwater tools were called, then ask “How many credits were used?” in that same conversation. A single prompt can call several tools. Being specific up front helps the assistant pick the right tool the first time.
My assistant doesn’t call Meltwater at all. Make sure the Meltwater connector is turned on in your AI assistant, then ask for something that needs Meltwater data, like the latest news about your brand. If the answer doesn’t show a Meltwater tool call, name Meltwater in your prompt.
A tool isn’t available to me. The tools you can use depend on your Meltwater package, and some need a permission an Account Admin can grant. Ask your assistant “Which Meltwater tools do I have, and what do they cost?” Checking costs doesn’t use credits. The Tools Explorer in the Developer Portal also lists your tools.