Glossary
Tool call
Tool call is defined here as follows. A function an MCP server exposes for a model to execute. The MCP specification treats a tool as code execution and says hosts must get explicit user consent before invoking one. The directory source for this term is Model Context Protocol specification.
Last verified 2026-10-17
Sources
Listings
Directory listings whose sources describe this mechanism.
| Name | Status | Category | Open source | Mechanism |
|---|---|---|---|---|
| Claude Agent SDK | Live | Frameworks and runtimes | No | Library that runs the Claude Code agent loop in a process you operate, with built in tools, permissions, sessions, hooks, and subagents. orchestration, tools, permissions |
| Google ADK | Live | Frameworks and runtimes | Yes | Open source framework for building, debugging, and deploying agents, including graph workflows and multi agent orchestration. orchestration, multi-agent, tools |
| Lakera Guard | Live | Governance and mandates | No | Guard API that screens an agent conversation in one call, including tool definitions, tool calls, and tool results, against a project policy. guardrails, tool-calls, policy |
| Model Context Protocol | Live | Protocols | Yes | Open protocol that connects LLM applications to external tools, resources, and prompts through a standard request interface. tools, resources, prompts |
| NeMo Guardrails | Live | Governance and mandates | Yes | Open source toolkit for programmable rails that block, alter, or validate user inputs, model responses, and tool calls. guardrails, policy, tool-calls |
| OpenAI Agents SDK | Live | Frameworks and runtimes | Yes | Library that runs an agent loop in your application, with tools, handoffs, guardrails, sessions, and human approval before a tool runs. orchestration, tools, approvals |
| smolagents | Live | Frameworks and runtimes | Yes | Open source Python library whose code agents write actions as code, with an option to run that code in a sandbox. python, tools, code-agents |
| Strands Agents | Live | Frameworks and runtimes | Yes | Open source SDK for a model driven agent loop in Python and TypeScript, with hooks, execution limits, and guardrails around each step. orchestration, guardrails, tools |
Questions
- What is Tool call?
- A function an MCP server exposes for a model to execute. The MCP specification treats a tool as code execution and says hosts must get explicit user consent before invoking one.
- Which listings use Tool call?
- Claude Agent SDK, Google ADK, Lakera Guard, Model Context Protocol, NeMo Guardrails, OpenAI Agents SDK, smolagents, Strands Agents.
- When was Tool call last verified?
- Last verified 2026-10-17.