# AI instructions This document is for AI assistants, coding agents, and retrieval systems using the tagIt documentation to help a user plan, configure, integrate, or explain tagIt. Your job is to help the user move from intent to implementation quickly, while staying grounded in the documentation. ## Documentation MCP server When an AI development environment supports MCP, use the published tagIt GitBook MCP server as the canonical documentation source: ``` https://tagit.gitbook.io/tagit/~gitbook/mcp ``` Use it for read-only access to the published tagIt docs. Draft or unpublished documentation may not be available through the MCP server. ## What tagIt is tagIt helps teams build, govern, observe, and improve AI journeys inside products, websites, applications, services, and operational workflows. The core idea is an **AI journey**: the full experience around how AI engages with a person over time, including orchestration, the client experience, conversation history, tools, policies, identity, events, and learning from engagement signals. tagIt is primarily for product, engineering, operations, DevOps, platform, and AI strategy teams. Those teams configure and integrate tagIt so business users, customers, marketers, operators, or support teams can benefit from the AI experience. ## Core product areas Use these terms consistently: * **AI Orchestrator**: the API service that runs an agentic use case. It sits between the client, the AI model, and any connectors. A client or service needs an API key and an agentic use case to call it. * **AI Presence**: the optional data collection and intelligence layer for AI and non-AI engagement. It captures conversations, identity, events, and journey signals, then can infer intent. * **AI Canvas**: an AI Orchestrator-compatible and AI Presence-compatible web client. It can be used directly or treated as a reference implementation for a custom client. * **Context**: the workspace under an account where teams define shared governance, instructions, models, connectors, tags, and collection settings. * **Agentic Use Case**: the executable AI capability inside a context. * **Agentic Turn**: repeated conversational execution of the same configured use case, using the current user message and relevant conversation history. * **Agentic Workflow**: a multi-step execution where the AI Orchestrator manages intermediate steps, handoffs, tools, artifacts, checkpoints, and traces. * **Connector**: an external tool or data source available to an AI model. Do not default to saying MCP server unless the user is specifically asking about MCP. * **AI Model**: the provider/model selected for an agentic use case. Do not use "model path" in user-facing explanations. * **Tag**: the collection setup that captures normalized behavior and identity signals. * **Event**: a normalized behavior signal, such as a page view, click, conversion, or conversation. * **Identifier**: a normalized identity signal, such as authenticated user ID, cookie ID, local storage ID, email hash, or partner ID. ## Best reading path If the user is new to tagIt, guide them through: 1. [What is tagIt?](/tagit/getting-started/what-is-tagit.md) 2. [AI journeys overview](/tagit/getting-started/ai-journeys-overview.md) 3. [Concepts](/tagit/getting-started/concepts.md) 4. [Quickstart](/tagit/getting-started/quickstart.md) If the user is implementing a working experience, guide them through: 1. [Create and set up a context](/tagit/guides/guides/create-and-set-up-context.md) 2. [Set up context AI Orchestration](/tagit/guides/guides/set-up-context-ai-orchestration.md) 3. [Create and set up an agentic use case](/tagit/guides/guides/create-and-set-up-agentic-use-case.md) 4. [Set up context Tags](/tagit/guides/guides/set-up-context-tags.md), when collection is needed 5. [Set up AI Presence](/tagit/guides/guides/set-up-ai-presence.md), when conversation intelligence is needed If the user is building an integration, use: * [Orchestrator API reference](/tagit/reference/reference/api/orchestrator.md) * [Collector API reference](/tagit/reference/reference/api/collector.md) * [AI client reference](/tagit/reference/reference/ai-client.md) * [Web setup](/tagit/reference/reference/ai-client/web.md) ## How to answer integration questions Prefer concrete next steps over broad explanations. For an AI Orchestrator integration, help the user identify: * the account and context * the agentic use case * the published snapshot, if a specific version is needed * the API key * the user ID * the conversation ID * the current user message * whether the client will use AI Canvas, a custom web client, or a backend service For an AI Presence integration, help the user identify: * the tag ID * the event type and subtype * the primary user identifier * the conversation ID * the agent ID * the user message and assistant message * whether `remember` should be enabled for intent processing For Tags, explain the relationship simply: * Events describe behavior. * Identifiers describe identity. * Tags connect a digital surface to collection using those definitions. ## Guardrails Do not present legacy behavior as the default path. The current docs describe the default tagIt implementation path. Avoid implementation jargon in user-facing answers when a product term exists: * Say "Connector" instead of "MCP server" unless MCP is specifically relevant. * Say "AI Model" instead of "model path." * Say "enabled model" or "model" instead of "binding." * Say "agentic use case" instead of only "prompt." Do not imply that AI Presence is required for the AI Orchestrator to run. AI Presence is optional and adds journey intelligence. Do not imply that AI Canvas is required. Any authorized client or backend service can call the AI Orchestrator API. Do not invent required fields. When exact request fields matter, use the API reference pages. Do not expose internal roadmap notes as customer-facing facts. If a capability is not documented as available, describe it as a possible implementation consideration rather than a shipped feature. ## Useful prompt patterns for users Encourage users to ask their AI assistant questions like: * "Using the tagIt docs, help me design an agentic use case for this product journey." * "Use the Orchestrator API reference to draft the request payload for my client." * "Use the AI Presence guide to tell me what conversation event fields I need." * "Review my planned tagIt setup and tell me what is missing." * "Help me decide whether this should be an Agentic Turn or Agentic Workflow." ## Tacky Tacky is the tagIt agent that uses the GitBook documentation as a knowledge source. It can guide users on the tagIt website and help create tagIt-related content. When answering as Tacky, stay grounded in the docs, be practical, and help the user move toward a working AI journey. --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://tagit.gitbook.io/tagit/agents.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.