Skip to content

Guide

How to let an AI agent manage your social media safely

Letting an agent post for you is useful right up to the post you did not mean to publish. This is how SPREVA keeps the access, decides what needs your yes, and keeps a record of everything the agent did.

The short answer

Connect the agent to SPREVA, not to your networks. SPREVA holds the access to your accounts and the agent never sees it. Every action has a risk level, and the workspace decides for each level whether the agent goes ahead, shows you a preview first, or waits for a person to approve. SPREVA enforces it on its own servers, whatever the agent's app does.

What can go wrong

Most of it comes down to an agent acting on a guess:

  • Publishing now when you meant to schedule.
  • The right post on the wrong account.
  • The same post twice, when the app repeats a call after a timeout.
  • Spending your AI credits on something you did not ask for.

What the agent never sees

From the day it connects:

  • Your password. You sign in on SPREVA's own page, and the app only receives access to the workspace you chose.
  • Your networks' access. What SPREVA holds for Instagram, TikTok and the rest is never part of an answer.
  • Other workspaces. A connection reaches only the workspace you allowed.
  • Anything after you disconnect it. Removing it under Settings, then Agents, ends its access.

Risk levels

Every call that changes something has one of three levels:

LevelWhat it coversUnless you change it
LowReading: search, overview, analytics, reviewing a videoAlways allowed
MediumDrafts, edits, scheduling at least an hour ahead, files, campaigns, queues, Studio projects and exportsAllow
HighPublishing now, scheduling less than an hour ahead, deleting a published post, spending AI credits on a voice-over or an imageAsk first

See Permissions and safety for every rule, and Tools for the level of each tool.

Allow, Ask first or Needs approval

  • Allow: the call goes ahead.
  • Ask first: the first call changes nothing and returns a preview of exactly what would happen: which accounts, which files, which caption and when. The agent shows it to you, and only a second call carries out that preview.
  • Needs approval: the call returns a link, and a person with the right role approves or rejects it in SPREVA. It is the only mode that proves a person saw it.

A repeated call with the same request id returns the first answer instead of doing the work again, so a retry never makes a second post or a second charge.

Set it up

  1. 1

    Copy the SPREVA connector address and add it to your agent:

    URL
    https://app.spreva.ai/api/mcp
  2. 2

    Sign in to SPREVA, choose the workspace and click Allow.

  3. 3

    In SPREVA, open Settings, then Agents, and choose a mode for medium and for high.

  4. 4

    Ask the agent for something real, and watch what it shows you before it acts:

Ask your agent
Publish the launch post on Instagram and LinkedIn now.

Questions

Does the agent get my Instagram or TikTok password?

No. You connect each network to SPREVA with that network's own sign-in, and the agent only talks to SPREVA. It never receives a password or an access token.

What if the agent picks the wrong account?

With Ask first, publishing now shows you the accounts before anything happens. And tools take names: a name that matches more than one account comes back as a choice, and nothing is saved until one is chosen.

How do I switch an agent off?

Under Settings, then Agents, remove it. Its access ends at once.

Can someone else approve what the agent does?

Yes. With Needs approval, the agent's request waits for a person with the right role to approve or reject it in SPREVA.

Can I see what the agent did?

Yes. Settings, then Agents, lists each connected app, the person it acts for and what it did recently.

From your next idea to a published video.