Adopt Glossary

Coordinated execution of multiple APIs by an AI agent.

Tool for defining agent actions using natural language prompts.

Structured format to define how agents interpret tasks.

A low-code platform for creating AI agents with built-in orchestration, reasoning tools, and API access—no complex coding needed.

Create, Read, Update, Delete operations performed by AI agents via API calls.

The memory span of an LLM, defining how much previous text it can use to understand and generate contextually relevant responses.

A technique where LLMs trigger external functions or tools by generating structured calls, enabling real-world task execution.

A unified layer for natural language interaction inside apps.

Model Context Protocol for standardized tool access across AI agents and APIs.

When an AI agent breaks complex goals into sequential steps, executing them one by one across tools or systems.

Environment to test and refine agent prompts before deployment.

The integrated stack of APIs, functions, and utilities that lets AI agents act within software products effectively.

A database optimized for similarity search using embeddings, key in agent memory.

An autonomous system powered by LLMs that can reason, take actions, and complete tasks across digital tools without human help.

Ensuring agent decisions are aligned with app logic and rules.

A new paradigm where software can reason, act, and adapt autonomously—powered by LLMs and dynamic orchestration layers.

Multi-step agent workflows that operate without manual intervention.

A reasoning method where the AI breaks problems into steps to improve accuracy, transparency, and multi-hop task execution.

User experiences designed around natural language dialogues.

A personalized AI assistant embedded into software workflows that helps users complete tasks faster using natural language commands.

The process of adapting a pre-trained LLM to specific domains or tasks by training it further on curated, relevant data.

A class of LLMs that generates human-like text using transformer architecture and massive pre-training datasets.

Advanced AI systems trained on massive datasets to understand and generate natural language across diverse domains.

Agents capable of understanding and processing text, images, voice, and other inputs.

Tasks triggered through human language commands within applications.

Linking multiple prompts together for complex, multi-turn tasks.

Retrieval-Augmented Generation enhances agent answers with real-time document fetches.

A training method where models learn through feedback and rewards, optimizing for long-term outcomes and behavior.

System that categorizes user prompts to assign correct actions.

Central interface to monitor and debug AI/agent performance.

Time delay between prompt and action execution in agents.

Identifying when agent behavior deviates from intended logic.

Detailed records of agent interactions and API usage.

Track every decision and API call your agent makes with structured logs for debugging, observability, and continuous optimization.

Dedicated environment to validate agent logic before going live.

Metrics that track how often and how well agents are used.

Backup plans when agent fails or lacks required context.

Defines the exact limits of what an agent can do per user.

Logs that capture every action and decision made by agents.

Ensures all agent data remains within required geographic regions.

Real-time masking of sensitive data in agent outputs.

A method for keeping AI decisions aligned with human judgment by adding approval steps, reviews, or guardrails.

Automatically hiding sensitive user data from logs and agents.

Role-Based Access Control for safe agent action execution.

Resolving the correct meaning of ambiguous terms or instructions using context, logic, and user-specific information.

Ensuring agents only execute actions users are authorized for.