> For the complete documentation index, see [llms.txt](https://docs.developers.symphony.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.developers.symphony.com/ai/ai-agent.md). # AI agent ## AI agents AI agents are specialized service accounts that bring LLM-powered reasoning into Symphony. They can answer questions, use Symphony chat context, and take actions on behalf of a user within the platform's security model. AI agents are accessible in a dedicated private channel called the **AI Agent stream**. ### Which AI agents can be found in Symphony? Several kind of AI agents can be found in Symphony: * **Built-in agent** — Symphony's built-in assistant for messaging tasks such as summaries and catch-ups. * **Custom agents** — Agents tailored to your users, tools, and workflows. You can create them in [AI agent studio](/ai/ai-agent/ai-agent-studio.md) or build them with Symphony APIs. * **Community agents** — Agents published by Symphony partners or the wider Symphony ecosystem. ### How AI agents differ from bots AI agents extend the standard bot model with user-scoped context and AI-specific behaviors. They can: * Reason over prompts and conversation context. * Use tools and workflows on behalf of a user. * Stream responses as they are generated. * Expose predefined prompts for repeatable tasks. ### Interaction modes AI agents support different interaction patterns depending on the task. #### AI Agent stream An **AI Agent stream** is a private, encrypted 1:1 conversation between a user and an AI agent. No other user can join. Use this stream when the agent needs a private working space, personal context, or a persistent session with the user. Because the stream is private, the agent can safely return user-specific results and perform user-specific actions there. #### Side panel The side panel lets a user chat with an AI agent while keeping the current conversation in view. The parent chat is passed as context to the agent. This is useful for: * Summarizing a room. * Drafting a reply from current context. * Triggering workflows based on the active conversation.
AI agent side panel

The side panel keeps the current conversation in context while the user chats privately with the agent.

### Context awareness AI agents can use a user's chat context to produce grounded, source-aware answers without joining the room itself. This allows the agent to help in internal rooms, external rooms, and direct messages, while keeping the interaction private to the requesting user. Symphony Messaging chat context can be shared with the agent in two main ways: * Implicitly, through the side panel. * Explicitly, through prompt message injection functions. {% hint style="warning" %} AI agents use **on-behalf-of APIs**. An agent can never access more data than the requesting user can access. Room memberships, information barriers, and sharing restrictions are enforced by design. {% endhint %} ### Core capabilities #### Predefined prompts AI agents can expose predefined prompts for common tasks. These prompts give users a fast way to run proven instructions without writing them from scratch. Use predefined prompts to: * Inject additional context into a request. * Standardize complex workflows. * Guide tool use across multiple steps. #### Streaming AI agents can stream responses as they are generated. This reduces time to first output and improves the user experience for longer answers. Each update still passes through Symphony compliance controls before it is shown to the user. #### Thinking AI agents can return their reasoning trace in the conversation when that experience is enabled. This allows: * users understand why the agent responded a certain way. * designers inspect tool calls and parameters. * teams improve prompts and edge-case handling. #### Per-user enablement An AI agent combines two Symphony assets: * A **bot** identity, which represents the Service Account. * An **app**, which controls the actions the agent can perform on behalf of a user. To message an AI agent, or create a new AI Agent stream, the linked app must be installed for that user. This gives administrators fine-grained deployment control. They can enable specific agents only for users who need them. ### Common use cases AI agents are useful for tasks such as: * Summarizing long discussions and catch-ups. * Drafting responses from the current chat context. * Running guided workflows with tool calls and user context. ### Create your first AI agent {% hint style="warning" %} AI Agent is an add-on to Symphony Services and may require additional licensing. Contact to enable it in your environment before you begin. {% endhint %} Choose the path that fits your team: * [AI agent studio](/ai/ai-agent/ai-agent-studio.md) — the fastest no-code path to create, configure, and manage agents. * Develop custom AI Agent *(coming soon)* — an API-first path for teams building custom agent logic and integrations. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## 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, and the optional `goal` query parameter: ``` GET https://docs.developers.symphony.com/ai/ai-agent.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. 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.