Every result in MCPFlo carries a Tokens tab, showing what that specific call costs in context-window terms — separate from the context-budget estimate for the tool’s definition (see Server Detail View & Token Estimator).

What it measures

The Tokens tab on a given result breaks down two figures:

• Response cost — how many tokens the content returned by this call would consume if fed into a model’s context (text, JSON, or any other content block in the result).
• Definition cost — a reminder of what the tool’s own schema/description costs just by being connected, shown alongside the response so you can see both sides of the ledger for that capability in one place.

Why per-call, not just per-tool

A tool’s definition cost is fixed — it’s paid once, just by connecting to the server. But its response cost varies every time you call it, depending on the arguments you pass and what the server returns. A search tool called with a broad query might return a huge result; the same tool called with a narrow one might return almost nothing. The per-call footprint lets you see that variance directly, call by call.

Why it’s useful

• Sizing responses before production use — if you’re about to wire a tool into an agent, check whether a typical call returns a response that fits comfortably in your context budget, or whether it needs pagination/truncation on the server side.
• Debugging bloated responses — if an agent using this server is burning through context unexpectedly, replay the same call in MCPFlo (see Replaying Past Calls) and check its token footprint directly.
• Comparing arguments — try the same tool with different inputs to see how the response size (and cost) changes.

How it’s calculated

Token counts are estimated using gpt-tokenizer. Since MCPFlo has no model in the loop, this is always an estimate — actual token usage may differ slightly depending on the specific model and tokenizer used in production.

Related

• See Understanding Responses for how the underlying content is rendered.
• See Server Detail View & Token Estimator for the fixed, per-server definition cost this complements.