Editing Portal:Artificial Intelligence (section)

=== MCP and AI tool calling ===

The '''Model Context Protocol''' or '''MCP''' became important in the mid-2020s as a way for AI assistants to interact with tools, files, APIs, and external systems. For Perl, MCP is especially relevant because Perl has always been strong at writing system tools and glue code.

Langertha includes MCP tool-calling support. Its `Langertha::Role::Tools` documentation describes a tool-calling loop that gathers available MCP tools, sends tool definitions to the LLM, executes returned tool calls, feeds results back to the LLM, and repeats until a final text response is returned.<ref>MetaCPAN, “Langertha::Role::Tools,” https://metacpan.org/pod/Langertha::Role::Tools, accessed July 3, 2026.</ref>

Perl also has domain-specific MCP work, such as `MCP::K8s`, which describes itself as an MCP server that gives AI assistants access to Kubernetes clusters with RBAC-aware dynamic tools.<ref>MetaCPAN, “MCP::K8s,” https://metacpan.org/pod/MCP::K8s, accessed July 3, 2026.</ref>

This is a natural area for Perl. MCP servers often need to wrap existing command-line tools, system APIs, databases, logs, and administrative interfaces. Perl is well suited to that kind of integration.