The persona connector shapes how your flespi AI agent presents itself and whom it serves. Give the agent a name of your own, describe its audience, decide whether it speaks under your brand alone, and hand it standing instructions in your own words. The agent then faces your customers or colleagues as your assistant, in their vocabulary, with every other capability intact.
New to flespi AI agents? Read the flespi AI agents article first: what agents are, how they reason and sleep, and how approvals keep you in control. This article covers the persona side of the story.
1. What a persona changes
Without a persona the agent is the flespi agent: it introduces itself as such, speaks about flespi items by their names, and treats the account owner as the person it serves. A persona replaces that presentation layer and nothing else. The tools, the connected systems, the knowledge, the approval flow and the conversation history all stay as they are.
The persona holds no credentials and queues no operations, so it has no approval rules of its own. One persona item can be assigned to any number of agents, and a change to it reaches every agent from its next reasoning step. Assign it alongside the connector of the system the agent operates: a Wialon Hosting account, a Teltonika FOTA WEB account, or the flespi account itself.
2. Configure the persona
| Field | What it does |
| name | How the agent presents itself, with your organization when wanted: "Acme Assistant", "Fleet Helper by Acme". Outbound messages carry it as the author name. |
| audience | Who the agent serves and what they do: "fleet managers of Acme's clients running their vehicles in the Acme tracking app". The agent reads requests, the connected systems and the documentation through that audience's vocabulary and answers in it. |
| disclosure | open (the default): the agent names flespi and the connected systems freely. branded: the persona's product is the only platform the audience hears about. Nothing the agent sends names flespi, its items, documentation or panel, and asked what it runs on it answers as the persona and leaves the technology behind it undescribed. |
| instructions | Standing guidance from you, in any language: the scope the agent takes on, what it declines, where it escalates, the terms the audience expects. The agent serves the audience's requests within it. |
The easiest way to write a persona is to ask your flespi agent: describe who the assistant is for and what it should and should not do, and the agent drafts the persona with you, creates the connector under AI -> Connectors (see how agents and connectors are created) and assigns it to the agents you name. You can equally fill the form by hand.
Assign the persona to the agent on its Connectors tab, next to the connector of the system it operates:
Two limits are worth knowing before you rely on branded disclosure. The boundary is held by the AI model following its instructions, not by a filter, so an application facing the public should keep its own check on the way out. And a file or page the agent publishes for the audience is delivered as a link on the flespi CDN, so its address shows where it is hosted. Agents on the economy intelligence level hold the brand boundary less reliably than the standard and advanced levels.
3. Serve your own users
A persona is what turns the agent into an assistant for the people you serve: a telematics platform giving its service providers an assistant that administers their accounts, or an integrator giving each fleet client an assistant inside its own application. The pattern is the same in both cases:
- one agent per customer, created from your backend through the agents API, with a connector holding that customer's credentials: their platform token, or a flespi token scoped to their subaccount.
- one persona shared by all these agents, carrying your name, your audience and your instructions.
- your application exchanges messages with each agent through its conversation, follows the live history over MQTT, and either shows the pending approvals to the customer or sets the operations ACL of the agent in advance.
The customer talks to the agent as the principal, with the full capability of the connected system behind it: reads run freely, every change waits for the approval rule you configured. You remain the owner: the agent, its connectors, its rules and its spending stay in your account.
Your application does not even need a chat of its own: HelpBox, the flespi chat application, embeds the agent chat by link, with a token limited to that one agent (see Talking to your agent). This is what the customer sees:
4. What the user sees
Say a tracking platform assigns a persona named "Acme Assistant" with branded disclosure to an agent connected to a service provider's account, and the provider asks who they are talking to. The agent introduces itself as Acme Assistant, lists what it can do for that account in the platform's own terms, and declines to describe the technology behind it. Asked for a report, it designs one in the platform's vocabulary, asks for the one decision that matters, and queues the change for approval. The same agent with open disclosure answers the same questions naming flespi where it is relevant.
The persona connector is one of several an agent can hold: browse the full connector list to give your agent more capabilities.
Subscribe to the flespi AI and MCP change log to stay in sync with any updates.