Skip to content
Picasi Picasi Picasi Documentation
Open App

Using Picasi in the AI assistant - Getting started

Imagine you’re working on a competitive analysis in Claude and want to know what Muster AG posted on LinkedIn last week. Instead of opening the Picasi app, you simply ask, “Show me the latest updates from Muster AG in Picasi.” Claude responds immediately with the current data.

This is possible via MCP—the Model Context Protocol. Picasi implements an MCP server that allows AI assistants to work with your Workspace data: search for updates, list sources, retrieve reports.

This lesson shows you how to get started with Claude Code, the most commonly used method.

What you need

  • A Picasi account with Admin or Owner role
  • Claude Code CLI (installed and logged in)

How to proceed

  1. Navigate to AI Connections

    In Picasi, open the Settings menu and click on AI Connections and Tokens. Here you manage this Workspace’s API tokens. Your own OAuth connections live elsewhere, under Personal settings → Connections.

    AI connections and tokens settings page with empty token list and Create new token button

  2. Create an API token

    Click on Create New Token. Give it a name that clearly indicates its purpose—for example, Claude Code – Lena or Claude Code – Agentur.

    Select the appropriate role:

    • Member — can read updates and generate reports; sufficient for most use cases
    • Admin — can also create and edit sources and channels

    Click Create Token.

    Newly created API token with complete token string and copy button

  3. Copy MCP URL

    On the same page you’ll find your MCP URL. It is the same for every Workspace:

    https://picasi.app/mcp

    Which Workspace the assistant sees is decided by the token from step 2. A token belongs to exactly one Workspace.

  4. Configure Claude Code

    Open a terminal and add Picasi as an MCP server to Claude Code:

    Terminal window
    claude mcp add picasi \
      --transport http \
      --url https://picasi.app/mcp \
      --header "Authorization: Bearer DEIN_TOKEN"

    Replace DEIN_TOKEN with the token from step 2.

    Alternatively, you can enter the configuration directly into the .mcp.json in your project root:

    {
      "mcpServers": {
        "picasi": {
          "type": "http",
          "url": "https://picasi.app/mcp",
          "headers": {
            "Authorization": "Bearer DEIN_TOKEN"
          }
        }
      }
    }

  5. Test the connection

    Launch Claude Code and ask something like:

    What did Muster AG post in Picasi last week?

    Or:

    Show me my latest reports in Picasi.

    If the connection works, Claude will respond with real data from your Workspace.

    Claude Code Terminal with active Picasi MCP server in the /mcp overview

What you can now do with Picasi in the AI Assistant

  • Search for updates by source, time period, and topic
  • Count Updates instead of listing them, by source or by month
  • Save, drop and tag Updates with topic tags
  • Create reports, retrieve them and process them in the chat
  • List and manage sources and channels (with an admin token)

You can find a complete list of available tools in the MCP Tools Reference.

What’s next?

You’ve connected Picasi to Claude Code. From here, you can incorporate Picasi data into any analysis without switching between tools.

MCP-Tools - Overview All available MCP tools with parameters and examples.
Connect Claude.ai Link Picasi with Claude.ai via OAuth.