vocabulary: name: Toolhouse Vocabulary description: >- Domain vocabulary and taxonomy for the Toolhouse AI agent platform. Covers agent lifecycle management, tool ecosystem, execution model, and platform concepts used across the Toolhouse Platform API. provider: Toolhouse url: https://toolhouse.ai version: "1.1" created: "2026-05-03" modified: "2026-05-22" terms: # Core Platform Concepts - term: AI Worker definition: >- Toolhouse's primary product noun (as of 2026): "a system that carries out a task with three components: a trigger, a process that may include specialized skills, and tools or systems it can connect to." AI Workers are designed for repeatable workflows and are used interchangeably with "AI Agents" in Toolhouse documentation. Every Worker can be invoked as an API via the Workers API surface at agents.toolhouse.ai. tags: - Core - AI Workers - AI Agents relatedTerms: - Agent - Workers API - Trigger - Integration - term: Workers API definition: >- The execution surface that lets any deployed Toolhouse worker (agent) be invoked over HTTP. Hosted at agents.toolhouse.ai. Supports text and NDJSON streaming responses, conversation continuation via the X-Toolhouse-Run-ID header, and stateful history retrieval via GET on the agent + run ID path. Authentication uses HTTPBearer with a Toolhouse-issued API key. tags: - Core - Execution - Streaming relatedTerms: - AI Worker - Run ID - Streaming - term: Run ID definition: >- The session identifier for a worker invocation, returned in the X-Toolhouse-Run-ID response header on the initial POST and used in subsequent PUT (continue) and GET (history) requests against /{agent_id}/{run_id}. tags: - Execution - Identity relatedTerms: - Workers API - Agent Run - term: Agent definition: >- An AI-powered application defined in plain language (Agent Editor) or as code, deployed to Toolhouse, and addressable via the Workers API. In current Toolhouse documentation the terms "Agent" and "AI Worker" are used interchangeably; "Worker" is the preferred product noun. tags: - Core - AI Agents relatedTerms: - AI Worker - Agent Run - Agent Studio - term: Agent Run definition: >- A single execution instance of a deployed Toolhouse agent. Captured as a record with unique run ID, timestamps, status, execution logs, and MCP server interaction logs. tags: - Core - Execution relatedTerms: - Agent - MCP Server Log - Log - term: Agent Studio definition: >- Toolhouse's visual, browser-based environment for interactively building and testing agents. Agent Studio chats are persistent sessions that capture conversation history. tags: - Development Tools relatedTerms: - Agent - Chat Session - term: Tool definition: >- A pre-built capability available in the Toolhouse tool library that agents can invoke. Tools include RAG memory, web search, code execution, browser automation, web scraping, database connectors, and communication tools like email and Slack. tags: - Core - Tools relatedTerms: - Bundle - Tool Execution - MCP Server - term: Bundle definition: >- A named, reusable collection of tools grouped together for a specific purpose. Bundles can be shared across multiple agents and published to the Toolhouse marketplace. tags: - Core - Tools relatedTerms: - Tool - Agent - term: MCP Server definition: >- The Toolhouse Model Context Protocol server that exposes the full tool library to any MCP-compatible AI client or agent framework. Allows external AI clients to call Toolhouse tools natively. tags: - Protocol - Integration relatedTerms: - Tool - MCP Server Log - term: MCP Server Log definition: >- A recorded log entry capturing interactions between an agent run and the Toolhouse MCP server, including tool calls, parameters, and results. tags: - Observability relatedTerms: - MCP Server - Agent Run - term: Schedule definition: >- A cron-based trigger configured on an agent that causes it to execute autonomously at specified times without manual invocation. tags: - Automation - Scheduling relatedTerms: - Agent - Cron Expression - term: Cron Expression definition: >- A standard five-field time expression (minute hour day month weekday) used to specify when a scheduled agent should execute automatically. tags: - Scheduling example: "0 9 * * 1-5 (every weekday at 9am)" relatedTerms: - Schedule - term: API Key definition: >- An authentication credential created in Toolhouse Settings that is passed as an HTTPBearer token to authorize access to the Platform API. tags: - Authentication - Security relatedTerms: - HTTPBearer - term: HTTPBearer definition: >- The authentication scheme used by the Toolhouse Platform API. API keys are passed in the Authorization header as "Bearer {token}". tags: - Authentication relatedTerms: - API Key - term: Agent File definition: >- A file (document, data, code) attached to an agent for use in RAG (retrieval augmented generation) or as contextual input. Files can be uploaded and managed per agent. tags: - Files - RAG relatedTerms: - Agent - RAG Memory - term: RAG Memory definition: >- A Toolhouse built-in tool providing Retrieval Augmented Generation capability. Enables agents to store and retrieve information from a persistent vector store across runs. tags: - Tools - Memory relatedTerms: - Tool - Agent File - term: Slug definition: >- A URL-friendly identifier automatically generated from an agent's name, used in public agent URLs (e.g., toolhouse.ai/agents/{slug}). Custom slugs can be set via the vanity slug endpoint. tags: - Identity relatedTerms: - Agent - term: Agent Subscription definition: >- A billing arrangement allowing users to access paid agents published by other Toolhouse users. Managed via Stripe checkout sessions. tags: - Billing - Marketplace relatedTerms: - Agent - term: Credit definition: >- Toolhouse's primary metered unit of consumption. One credit corresponds to one worker run on the Toolhouse runtime. Monthly credit allowances are set per plan (Business 25,000, Business Max 80,000, Enterprise volume-priced). Model token usage is not separately metered — every plan includes "free unlimited tokens." tags: - Billing - Metering relatedTerms: - Agent Run - Worker Email - term: Worker Email definition: >- A per-worker email inbox provided on every Toolhouse plan that lets external systems trigger workers by sending email to a dedicated address. tags: - Triggers - Integration relatedTerms: - AI Worker - Trigger - term: Trigger definition: >- The mechanism that initiates a worker — manual invocation, a Workers API call, a Schedule, a worker email inbox message, or a webhook from an integration such as Zapier or n8n. tags: - Triggers - Automation relatedTerms: - AI Worker - Schedule - term: MCP Discovery definition: >- A Toolhouse feature (introduced August 2025) that lets agents discover and attach to thousands of Model Context Protocol servers at runtime without per-server configuration. tags: - MCP - Discovery relatedTerms: - MCP Server - AI Worker - term: Virtual Computer definition: >- Toolhouse's isolated, internet-connected sandbox capability for running Python code as part of a worker. Provides per-run compute isolation for code execution and data processing. tags: - Tools - Sandbox relatedTerms: - Tool - Code Execution - term: Toolhouse RAG definition: >- Toolhouse's managed retrieval-augmented generation tool that ingests attached files and skills/knowledge into worker context without requiring the developer to operate a separate vector store. tags: - Tools - RAG relatedTerms: - RAG Memory - Agent File - term: Theme Preference definition: >- Visual customization settings for an agent's public-facing interface, including color scheme and branding. Configured via the theme endpoint. tags: - Customization relatedTerms: - Agent tags: - Core - AI Agents - AI Workers - Authentication - Automation - Billing - Customization - Development Tools - Discovery - Execution - Files - Identity - Integration - Marketplace - Memory - MCP - Metering - Observability - Protocol - RAG - Sandbox - Scheduling - Security - Streaming - Tiers - Tools - Triggers