Skip to main content
The following terms appear throughout the H Tech Hub, and within our SDK, APIs, and tools.
TermDefinition


Agent		Agents are configurable and customizable AI-powered systems that can execute tasks using web automation. An agent on the Agent Platform consists of both code and configuration. When combined, this produces an agent identifier that enables you to run the agent on the Agent Platform.
Agent RunA single instance of an agent being executed. This is tracked via a trajectory.
Agent IdentifierA unique value or number assigned to an agent based on code and configuration. Agent identifiers enable you to run the agent on the Agent Platform.
Agent CatalogueH’s list of available agents.
Agent UI StudioA suite of react UI components for building agent interfaces, installable via the shadcn cli using the shadcn registry code distribution system.
BenchmarkingThe process of measuring the performance and efficiency of the agents and models in our catalogue. Benchmarking results vary by agent or model.
Holo1.5An Action Vision-Language Model (VLM) developed by H Company for use in the Surfer-H web agent system. Holo1.5 is designed to interact with web interfaces like a human user.
TrajectoryRepresents a single execution of an agent, including its objective, configuration, event log, status, and other metadata.
Trajectory IDThe unique identifier for a given trajectory or agent run. Trajectory IDs can be used to fetch logs, event data, and more.
TaskThe goal or instruction you give the agent, such as “Check product availability” or “Find recent news about H Company.”
EventA discrete step or action the agent performs during a run, for example thoughts, clicks, screenshots, and answers. Events are recorded as part of a trajectory.
Event LogA chronologically ordered list of events captured during the execution of a trajectory. Event logs are useful for debugging or analyzing agent behavior.
Agent Platform SDK / AgP JavaScript SDKOur lightweight JavaScript library for building, running, and managing agents. It includes methods for selecting agents, managing trajectories, handling errors and more.
Agent Platform APIA set of endpoints that let you create and manage trajectories, list agents, submit feedback, and more.
live_view_urlA link for observing an agent’s behavior during or after a run.
agent_identifierA unique ID used to specify which agent to run on the Agent Platform.
agent_run_idAn identifier that can be used to track and identify specific runs in a trajectory using the Agent Platform API.
Web TaskA type of agent task that involves navigating and interacting with web pages.
⌘I