Connect

MCP

Phersa has a remote MCP server at https://mcp.phersa.com/mcp. Connect your AI agent to it and the agent can search proven videos, read their transcripts and breakdowns, build your Create and Edit projects, and set up your Comment & DM Automation for you.

Docs menu MCP

Connect your agent

  1. In your agent, add a custom MCP server or connector with this URL:

    https://mcp.phersa.com/mcp

  2. Sign in with your Phersa account when the agent asks.

  3. Ask your agent something like "Find a trending video about skincare in Phersa".

The agent setup pages have exact steps for Claude, ChatGPT, Cursor, Codex, Gemini CLI and more.

Permissions

When your agent signs in, it gets permission to read and to write in one step.

  • Read (discover:read, library:read): search the video library and see your videos, personas, projects, results and automations.
  • Write (generate:write): track accounts, save videos, set up personas and projects, generate, edit and export videos, manage automations, and delete any of these.

If your agent can read but not write, its connection is older than this: remove Phersa from your agent and add it again.

Spending

Generations bill to your own API keys. When your agent asks for a generation, Phersa first returns a plan and spends nothing. The generation only runs when the agent sends it again as real, after you confirm.

Comment & DM Automation

Your agent can list your connected accounts and DMs left, set up and edit Comment & DM Automation on your posts, and read who commented and whether their DM went out. Connect Instagram or Facebook in Settings → Socials first: your agent can't connect or remove accounts, buy add-ons or change your plan. Instagram DMs are text only, so the link goes in the text; Facebook DMs can have up to 3 link buttons.

The same persona in every clip

Ask your agent to make the images chained: each clip's image then starts from your persona's photo and the image of the clip before, so the face and the scene carry through. Your agent asks whether you want to check each image first (it makes one clip, you keep it or redo it, then the next) or have them all made in order. The references it uses show on each clip in Create, as if you had added them. Each video then starts from its own clip's image.

Tools

40 tools, grouped like the app's tabs. The MCP page lists every tool with its permission.

GroupTools
Discoverdiscover-search video-get
Tracktrack-account-list track-account-add track-account-remove
Collectcollect-list collect-add collect-delete
Personaspersonas-list personas-voices-list personas-write personas-delete
Createcreate-project-list create-project create-project-delete create-clip create-clip-delete create-models-list create-settings create-settings-set settings-keys-status
Jobsjobs-create jobs-get
Assetsassets-list assets-add assets-delete
Resultsresults-list results-delete
Editedit-project-list edit-project edit-project-delete edit-media edit-clip edit-caption edit-track
Monetizemonetize-list monetize-posts monetize-activity monetize-automation monetize-automation-delete

Start from a video

On any video in Discover, Track or Collect, choose MCP to copy a prompt that has your agent open that video with its transcript and breakdown.