# `MCP002` Multiple queries per session
{bdg-ref-info}`mcp | core <mcp:core>`
{bdg-ref-info}`mcp | performance <mcp:performance>`
> I would like to send multiple queries in one conversation without the overhead of
> reloading the specification files each time. When exploring a specification interactively, I might
> search for tags, then get specific requirements, then check their tests - all in quick
> succession. Fast responses are crucial for a smooth user experience.
> 

The system shall be able to handle multiple tool calls in a single session
without restarting.

The server shall keep the specifications loaded in memory and respond to each query
without reloading the YAML files.

This allows efficient querying during interactive sessions with an LLM agent.

Queries are received by the server in its standard input,
and responses are written to its standard output.


:::{dropdown} Properties
:icon: note
:color: primary

__author__: Claude

__since__: `0.2.0`

:::
:::{dropdown} Tested by
:open: 
:icon: check-circle-fill
:color: success

[`TMCP003` Initialize the MCP server](/tests/TMCP003)
:::

:::{dropdown} Source
:icon: file-code
:color: info

__Source file__: [`specs/mcp/nonfunctional.yaml`](https://github.com/agagniere/speky/blob/master/specs/mcp/nonfunctional.yaml)

__Loaded from__: [`speky_mcp`](https://github.com/agagniere/speky/blob/master/specs/mcp/mcp.toml)

__Implemented in:__
- {fab}`python` [`python/speky_mcp/server.py:1`](https://github.com/agagniere/speky/blob/master/python/speky_mcp/server.py#L1)
- {fab}`python` [`run_server`](https://github.com/agagniere/speky/blob/master/python/speky_mcp/server.py#L101)
:::