What AI agents actually do with your MCP server

Description and use

MCP servers appear healthy yet agents face hidden failures such as missing tools, argument mismatches, and inefficient responses. mcpspan revealed five issues that can be fixed to improve agent interactions.

Problem

MCP servers appear healthy but agents encounter hidden failures such as missing tools, argument mismatches, and inefficient responses.

Outcome

After instrumentation, five hidden failure patterns were uncovered and addressed, leading to clearer agent interactions and reduced error rates.

Workflow

• Instrument the server with mcpspan
• Run simulated traffic
• Collect metrics on tool calls, arguments, and response sizes
• Analyze patterns of missing tools, argument errors, description issues, looping calls, and large responses
• Apply fixes: rename tools, adjust descriptions, add limits or summaries
• Re‑measure to confirm improvements

Evidence

Agents guess tool names. Some guesses are near misses: searchFlights when the tool is search_flights, seatMap instead of seat_map. Others are a tool the agent assumed should exist, like get_flight_price.

Source

When an agent sends arguments that don't match the input schema, the MCP SDK rejects them before your handler runs and sends the validation error back to the model as a normal result.

Source

The fix was one sentence in the description. The dotted line on the chart is where it changed: Errors drop right at the line, while the code didn't change at all.

Source

An agent tries check_in, gets "Check-in opens 24 hours before departure", and calls check_in again with exactly the same arguments.

Source

A tool that usually answers with 6 kB and sometimes with 650 kB looks fine on every latency chart, because it's fast. But that answer lands in the agent's context window, crowds out everything else, and costs tokens on every following turn.

Source