The MCP server for AI personas

Phersa's MCP server connects your AI agent to Phersa, the platform for building and monetizing AI personas. Add it to Claude, ChatGPT, Cursor or any MCP client, sign in with your Phersa account, and your agent gets 40 tools to search the library, manage personas, create videos and run comment-to-DM automations.

Add this URL to your agent

https://mcp.phersa.com/mcp
  • Sign in with OAuth, no API key to paste
  • 40 tools on one URL
  • Every generation shows its price before anything is spent
Setup steps for your AI agent
How it works

Connect your AI agent in three steps

One URL, one sign-in, then you ask for what you want.

  1. Add the server

    Paste the URL wherever your agent keeps its MCP servers. There is nothing to install or host.

  2. Sign in once

    Your agent opens Phersa's sign-in in your browser. Allow it once: that covers reading and making changes.

  3. Ask in plain words

    Your agent picks the tools itself, and nothing is generated until you confirm.

Tools

What your agent can do in Phersa

The same library, personas and projects as the app, on the same account.

Reference

All 40 Phersa MCP tools

Every tool, what it does and whether it only reads or makes changes. One sign-in covers all of them.

Tool What it does Access Permission
Discover
video-get Open a video from Discover, Track or Collect, with its transcript, analysis or similar videos when asked. Read discover:read
Track
track-account-list List the accounts you track, or the videos from one of them. Read library:read
track-account-add Track an Instagram, Facebook or TikTok account from its profile link. Write generate:write
track-account-remove Stop tracking an account. Its videos leave your Track feed. Write · can delete generate:write
Collect
collect-list List the videos you collected from links or uploads, newest first. Read library:read
collect-add Add a video from an Instagram, Facebook or TikTok link, or from an https video file. Write generate:write
collect-delete Delete a collected video. Write · can delete generate:write
Personas
personas-list List your personas, or open one with its generated results. Read library:read
personas-voices-list Search the voices in your ElevenLabs library. Read library:read
personas-write Create or update a persona: name, photo, world notes and voice. Write generate:write
personas-delete Delete a persona. Its generations stay in Results. Write · can delete generate:write
Create
create-project-list List Create projects, or open one with its full prompts, references and takes. Read library:read
create-project Start a new project, import clips from JSON, update its settings, or compile finished takes into a video. Write generate:write
create-project-delete Delete a Create project. Its generations stay in Results. Write · can delete generate:write
create-clip Add a clip, or change one clip's image prompt, video prompt, dialogue, models and references. Write generate:write
create-clip-delete Delete a clip. A project keeps at least one. Write · can delete generate:write
create-models-list List the image, video and voice-swap models you can use. Read library:read
create-settings Read your default models, aspect ratio and resolution. Read library:read
create-settings-set Change your default models, aspect ratio and resolution. Write generate:write
settings-keys-status See which generation providers you have connected, as yes or no only. Read library:read
Jobs
jobs-create Plan or start an image, video or voice generation, or a video export. It plans by default and spends nothing. Chained images keep the character consistent across clips: one clip at a time for you to check, or all in order. Write generate:write
jobs-get Check whether a generation or an export is still running, completed or failed. Read library:read
Assets
assets-list List your images, videos and audio, or open one file. Read library:read
assets-add Save an image, video or audio file from an https link to Assets. Write generate:write
assets-delete Delete one file from Assets. Write · can delete generate:write
Results
results-list List your generations and exported videos, newest first. Read library:read
results-delete Delete a result. Write · can delete generate:write
Edit
edit-project-list List Edit projects, or open one with its timeline. Read library:read
edit-project Start a new video, rename it, change its aspect ratio, or run Auto Captions or Auto Cut. Write · can delete generate:write
edit-project-delete Delete an Edit project. Its exports stay in Results. Write · can delete generate:write
edit-media Attach files to the media bin, or remove them. Write · can delete generate:write
edit-clip Add, update, remove or split a clip on the timeline. Write · can delete generate:write
edit-caption Add, update, remove or style caption text. Write · can delete generate:write
edit-track Hide or show a video, audio or caption track in the export. Write generate:write
Monetize
monetize-list List your connected accounts and DMs left this period, an account’s automations, or one automation with its comments, DMs sent and failed. Read library:read
monetize-posts List an account’s published posts, to pick the post an automation runs on. Read library:read
monetize-activity See who commented, what they wrote, and whether the DM and the public reply went out. Read library:read
monetize-automation Create or change an automation: keywords, DM versions, public reply, skip and hide words, delay, on or off. Write generate:write
monetize-automation-delete Delete an automation. Write · can delete generate:write
Permissions

Your agent only does what you allow

  • Signs in, no pasted key

    Your agent signs in through the browser and holds an OAuth token, so no Phersa secret sits in a config file.

  • One sign-in, read and write

    When your agent signs in, it gets read and write access together: it can search and read, and create, edit, track, collect, generate and change automations.

  • Plans before it spends

    jobs-create first returns a plan with the exact price and spends nothing. Your agent generates only after you agree, with an explicit dry_run=false.

  • Some things stay in the app

    Your agent manages Comment & DM Automation, but connecting or removing Instagram or Facebook accounts, buying add-ons, changing your plan and uploading files from your device happen only in the app. A file you drop into the chat can't reach Phersa: upload it in the app, or give your agent a public https link.

  • Automations send real DMs

    A new automation starts on and DMs the people who comment, from your account. Check the keywords and the message before your agent turns one on.

  • Your API keys stay private

    Provider API keys stay in Phersa Settings. The MCP server can't read or change them, and generation bills your own provider at its published rate. How API keys work.

  • Media links expire fast

    Links to videos and images that your agent receives stop working after about 2 minutes.

Connection

MCP server details

  • Endpoint: https://mcp.phersa.com/mcp · Streamable HTTP · OAuth 2.1 with PKCE. Metadata: https://mcp.phersa.com/.well-known/oauth-protected-resource/mcp. Dynamic client registration and client ID metadata documents are supported.
  • Scopes: discover:read and library:read for reading; generate:write for creating, editing, tracking, collecting, generating and changing automations. Write tools called without generate:write return "Missing scopes: generate:write".
  • 40 tools: discover-search and video-get for the library; track-*, collect-*, personas-*, create-*, edit-*, jobs-*, assets-*, results-* and monetize-* for the user's work.
  • jobs-create defaults to dry_run=true and returns a plan without spending. dry_run=false starts a real generation, billed by the user's own provider.
  • Provider API keys stay in Phersa Settings; the MCP server cannot read or change them.
  • Instagram and Facebook accounts are connected in Phersa Settings; the MCP server manages their comment-to-DM automations but cannot connect or remove accounts.
  • Buying add-ons, changing the plan and uploading files from the user's device happen only in the app; public https file links work with assets-add, personas-write and collect-add.
Agents

Works with the AI agent you already use

Each agent has its own page with the exact setup, or compare all agents to pick one.

MCP questions

What is the Phersa MCP server?

The Phersa MCP server lets an AI agent use Phersa for you. At https://mcp.phersa.com/mcp, it gives Claude, ChatGPT, Cursor and other MCP clients 40 tools to search proven viral videos, track accounts, collect videos, manage personas, build, generate and edit short videos, and run comment-to-DM automations on Instagram and Facebook.

Which AI agents work with the Phersa MCP server?

Any MCP client that supports Streamable HTTP and OAuth sign-in. Phersa has setup pages for many agents, including Claude Desktop, Claude Code, ChatGPT, Cursor, Codex, Gemini CLI and Grok.

See all agents →

Can my AI agent spend money in Phersa?

Only on your own provider API keys, and only when you agree. jobs-create plans by default, shows the exact price and spends nothing. It spends only after you say yes, and your provider bills you at its published rate, never Phersa.

Do I need a paid Phersa plan?

Yes. An account without a plan can connect, but its agent gets no tools and is asked to subscribe.

See pricing →

Can my AI agent see my API keys?

No. Provider API keys stay in Phersa Settings, and the MCP server can't read or change them. settings-keys-status only reports which providers are connected, as yes or no.

What does my agent need to generate images and video?

A paid Phersa plan and at least one provider API key connected in Phersa Settings, such as Google AI Studio, fal or Kie.

About API keys →

What is MCP?

MCP (Model Context Protocol) is an open standard for connecting AI agents to apps and data. An MCP server offers tools, and an agent such as Claude or Cursor calls them when a task needs them.