--- title: "Pioneer AI" description: "Pioneer AI is the assistant in the Outpost console. It reads your fleet through tools, proposes changes for your approval, and keeps conversations." url: "https://support.outpostplatform.com/pioneer-ai/" product: "depot" type: "concept" status: "draft" ai_summary: "Pioneer AI is the console assistant (ai:use). It reads endpoints, alerts, tickets, patches, scripts, and documentation with the user's permissions, proposes changes a technician approves, navigates the console, shows tables and charts, and offers remote desktop and shell cards. Conversations are kept encrypted for 30 days by default. Sources: Outpost Cloud AI or your own endpoint." availability: plans: ["home-lab", "essentials", "enterprise", "msp"] permission: "ai:use" preview: false source: "https://gitlab.com/outpostplatform/docs/-/edit/main/src/content/docs/pioneer-ai/index.md" license: "CC BY 4.0" --- # Pioneer AI Pioneer AI is the assistant built into the Outpost console. A technician asks a question in plain language, and Pioneer AI looks up the answer in your Outpost: which endpoints are offline, why a patch failed, what an alert means, which tickets are waiting. It can also take action, such as running a command or deploying software, after a technician approves it. This page explains what Pioneer AI can do, how it stays inside each person's permissions, and where its answers come from. ## What it can do Pioneer AI works through tools, the same set that [AI apps connected with MCP](/pioneer-ai/mcp-server/) use, plus a few that act on the console itself. | Task | Examples | | --- | --- | | Diagnose | Look up an endpoint's inventory, software, alerts, and patch state. Read command and script results. Search your documentation and knowledge base. | | Act on endpoints, device groups, and sites | Run a command or a script, open a shell, apply patches, deploy software, wake an endpoint, move endpoints to a site, change the members of a static device group. | | Work the service desk | Find, create, and update tickets, add messages, log time, and acknowledge alerts. | | Report | Build a table or a chart in the conversation from what it found, and reach any other console operation you have permission for, such as a report. | | Map the network | Read a client's network map, trace a path between devices, list what changed, compare a diagram with the network, and draw a new diagram. See [Network maps with Pioneer AI and MCP](/network-maps/pioneer-and-mcp/). | | Drive the console | Open a console page for you, such as the endpoint it is talking about. | | Hand you the controls | Show a **Connect** card for remote desktop or an **Open shell** card for a shell on an endpoint. The session starts only when you select the card. | | Work on an endpoint's screen | See and operate the screen of a remote session you have open, for one task you approve. See [Let Pioneer AI control a session](/remote-access/pioneer-control/). | | Research | Search the web for current versions, patches, and advisories, and cite its sources. See [Pioneer AI web search](/pioneer-ai/web-search/). Answer questions about Outpost from this documentation. See [Pioneer AI and the Outpost documentation](/pioneer-ai/outpost-docs/). | | Work on large tasks | Show a plan checklist, hand parts of a task to sub-agents, keep watching a long job after the reply ends, and undo changes that can be reversed exactly. See [Use Pioneer AI](/pioneer-ai/use-pioneer-ai/). | | Remember | Keep notes between conversations, for you, for one client, or for everyone. See [Manage Pioneer AI memories](/pioneer-ai/memories/). | Pioneer AI also knows which console page you are on, so "why is this one offline?" on an endpoint's page needs no further detail. See [Use Pioneer AI](/pioneer-ai/use-pioneer-ai/). ## How it stays inside your permissions - **It acts as you.** Every tool call runs with the permissions of the person chatting, under the same plan, [access rules](/depot/access-rules/), and client scope as their own console session. Pioneer AI cannot reach anything that person could not reach. - **It asks before it changes anything.** Looking things up runs without asking. Every change waits for an approval card in the conversation. See [Approve Pioneer AI actions](/pioneer-ai/approve-actions/). - **Some things are never allowed.** Pioneer AI cannot reveal secrets, play session recordings, export data, change billing, sign-in, or security settings, or change its own AI settings. See [Actions that are never allowed](/pioneer-ai/approve-actions/#actions-that-are-never-allowed). - **Everything is logged.** Every tool call appears under **Recent activity** on **Settings** > **AI**, with **Pioneer AI** in the **Via** column, and every change it makes appears in the [audit log](/depot/audit-log/). An owner can also turn on [automatic approval](/pioneer-ai/automatic-approval/), so that a separate AI safety review may approve some actions in place of the technician. ## Where the answers come from An owner chooses the source on **Settings** > **AI**: | | Outpost Cloud AI | Your own endpoint | | --- | --- | --- | | Who runs the model | An AI service provider Outpost contracts with | You, or a provider you choose | | Billed by Outpost | Yes, by use. Home Lab has a free monthly allowance instead. | No | | Budget in the console | A monthly budget | Organization and per-person budgets, from prices you enter | | Available on Remote Outpost | No | Yes | See [Set up Pioneer AI](/pioneer-ai/set-up/) to choose, and [Pioneer AI data and privacy](/pioneer-ai/data-and-privacy/) for what each source receives. ## Conversations are kept Each conversation is saved, encrypted, in your Outpost, so you can come back to it from the history list or open it in its own window. Conversations are deleted after the retention period an owner sets, 30 days after the last message by default. People who hold `ai:review`, which the owner holds by default, can read every technician's conversations. Conversations are never used to train models. See [Pioneer AI data and privacy](/pioneer-ai/data-and-privacy/). ## Permissions | Permission | Lets a person | Held by default | | --- | --- | --- | | `ai:use` | Chat with Pioneer AI, save personal prompts, and view **Settings** > **AI** | Every role | | `ai:manage` | Change **Settings** > **AI**, shared prompts, AI budgets, and a client's AI override, and view **Usage** | The owner. On an MSP's practice console, every role that holds `billing:manage`. | | `ai:review` | Read every technician's Pioneer AI conversations | The owner. On an MSP's practice console, every role that holds `ai:manage`. | | `remote:ai_control` | Let Pioneer AI control a remote session they have open, with `ai:use` | Every role that holds `remote:initiate` | See [Roles and permissions](/depot/roles-and-permissions/). ## How MSPs control it per client On the MSP plan, the Pioneer AI settings belong to your practice and apply to every client. A client policy turns Pioneer AI off for every client, off unless turned on per client, or on unless turned off per client, and each client's page can override it. A conversation started while a client is selected belongs to that client. If any client a conversation has worked with is turned off, the conversation stops with **A client this conversation used has AI access turned off by your organization's policy.** See [Choose which clients get Pioneer AI](/pioneer-ai/set-up/#choose-which-clients-get-pioneer-ai). ## Related - [Set up Pioneer AI](/pioneer-ai/set-up/) - [Use Pioneer AI](/pioneer-ai/use-pioneer-ai/) - [Approve Pioneer AI actions](/pioneer-ai/approve-actions/) - [Turn on automatic approval](/pioneer-ai/automatic-approval/) - [Pioneer AI web search](/pioneer-ai/web-search/) - [Pioneer AI and the Outpost documentation](/pioneer-ai/outpost-docs/) - [Manage Pioneer AI memories](/pioneer-ai/memories/) - [Give Pioneer AI instructions](/pioneer-ai/instructions/) - [Pioneer AI data and privacy](/pioneer-ai/data-and-privacy/) - [Review Pioneer AI usage and conversations](/pioneer-ai/usage/) - [Connect AI tools with MCP](/pioneer-ai/mcp-server/)