Ask your editor which product needs you

Setup

1. Sign in and connect at least one product. MCP access is part of Pro.
2. Open Settings, then API Keys, and create a key. Copy it: it is shown only once. Tick Allow write access (ad spend and markers) only if your agent should sync ad spend.
3. Paste one of the configurations below into your client.

claude mcp add --transport http plask https://plask.dev/api/mcp --header "Authorization: Bearer plk_YOUR_KEY"

{
  "mcpServers": {
    "plask": {
      "url": "https://plask.dev/api/mcp",
      "headers": {
        "Authorization": "Bearer plk_YOUR_KEY"
      }
    }
  }
}

https://plask.dev/api/mcp

curl -X POST https://plask.dev/api/mcp \
  -H "Authorization: Bearer plk_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_portfolio_pnl","arguments":{}}}'

The tools

Portfolio

get_portfolio_pnl

Revenue, ad spend, fixed costs and profit per product for a period, with the change versus the previous period and the date the data runs through. Missing sources come back as null, never as zero.

get_focus_cards

This week's Focus cards (at most three): the rule that fired, the product, the evidence numbers behind it, and the one-line explanation.

compare_products

Two or more products side by side over a period: revenue, profit, traffic, revenue per visitor, and trend direction.

Ad spend and markers, synced by your AI agent

Plask doesn't connect to ad platforms. Your agent reads spend from Meta, Apple Search Ads, Google Ads or TikTok with your own access, then syncs the daily totals here over MCP, and you're responsible for those platforms' terms. Write tools need a key created with write access. A sync replaces only the dates it sends, refuses days you entered by hand, and every batch can be undone under Ad spend on Sources & costs. For Apple Ads only the spend amount is stored, never impressions, clicks or installs. Synced spend shows in your P&L after the next refresh (today's spend by the next morning) and stays private: never on the board.

list_products

Your product ids and names, so the agent can map ad accounts to products.

sync_ad_spend

Write tool. Daily spend for one product, channel (meta, apple_search_ads, google_ads, tiktok_ads or other) and optional campaign, in any currency with an ECB rate. Other channels may add impressions, clicks, installs and conversions; for Apple Ads only the spend amount is accepted. Up to 400 rows (one per day) covering at most 400 days, no future dates, amounts from 0 to 1,000,000 a day. It replaces only the dates you send, so re-running never double counts, and if any day overlaps spend you entered by hand, it writes nothing and lists those days in the error. Each call is a batch, and the agent gets its batch ID back.

get_ad_spend

Reads back spend for a date range, both what you entered in the app and what your agent synced, with the batch ID of each synced row.

delete_ad_spend

Write tool. Removes synced spend for a product, channel, optional campaign and a date range of at most 400 days. Spend you entered in the app is never touched. Like every write, a delete is a batch you can undo.

undo_batch

Write tool. Undoes one write by its batch ID, including a delete: it restores what that batch replaced or deleted and removes what it added. If a later write changed the same days, undo that one first, and undo is refused if it would land on days you've since entered by hand. No time limit. The same Undo is under Ad spend on Sources & costs, in the list titled “Synced by your AI agent”.

create_marker

Write tool. Marks a release, price change or campaign start on a product's charts. Labels are plain text, up to 120 characters, with no links, and the same label on the same day isn't added twice. Each synced campaign also gets one marker, not one per day.

list_markers

Markers for one product or all of them, optionally within a date range.

Sources and products, set up by your AI agent

With a key you create with “Allow adding and organising revenue sources”, your agent can connect a revenue source and put its apps on the right products. It connects exactly the way the app does, with the same checks, and stores what you hand it encrypted. Credentials are never returned or logged; results show only a masked id. A key you give an AI agent also passes through the AI provider and the agent's own logs, which Plask doesn't control, so use a dedicated, read-only key you can revoke. You are responsible for the keys you give your agent. Names that come from your sources are returned to the agent marked as untrusted data, not instructions. Sources it adds show “Added by your AI agent via MCP” in Sources, and every change except creating a product or disconnecting can be undone with undo_batch.

list_connections

Your revenue sources with status, last synced day, a masked key id (never the key) and which products each one feeds.

list_unmapped

Apps, products, packages or properties on your connections that no product uses yet.

create_product

Write tool. Creates a product, within your plan's product limit.

add_source

Write tool. Connects Stripe, RevenueCat, Google Play or App Store Connect exactly as the app does: the same checks, stored encrypted, first sync started. Stripe takes only a restricted read-only key (rk_live_…), never a secret key; Google Play only a service account JSON key; App Store Connect only while you have accepted its terms in Sources (withdrawing them there stops your agent adding App Store Connect, while sources already connected keep syncing; a new terms version needs accepting again). Credentials you pass are stored encrypted and never returned. With a product, clear matches are mapped to it and the rest come back as choices; a product that already has an account of the same source gets a second one only if the agent explicitly allows a duplicate. At most 10 per hour per key.

map_to_product

Write tool. Puts an unmapped app, product or property on a product.

move_to_product

Write tool. Moves a mapped app, product or property to another product, history included.

disconnect_source

Write tool, only for sources your agent added; anything you set up in the app is disconnected in the app. It takes two steps: the first call returns a warning that it deletes the stored key irreversibly, and a confirming call within 5 minutes does it. Syncing stops and history stays. It can't be undone because the key is gone.

GA4 and alerts

list_properties

Lists your connected GA4 properties with their id, display name, platform (web, iOS, Android), and whether monitoring is enabled.

get_metrics

Daily rows for one property over the last 1 to 90 days: active users, sessions, new users, page views, revenue, transactions, session duration, engagement rate. Includes period totals and percentage change versus the preceding period.

get_alerts

Recent anomaly alerts with metric, date, expected vs actual value, modified Z-score, severity, direction, priority score, and the AI root-cause explanation when available. Filter by property and read state.

get_digest

The latest digest: the summary plus the metric snapshot it was generated from.

get_insights

Latest proactive insights: trends, milestones, personal bests, and pattern shifts.

get_portfolio_summary

Every enabled property with last-7-day totals, change versus the previous 7 days, and unread alert count, sorted by sessions.