The Abacum MCP connector is no longer admin-only. Anyone on your team can now connect Abacum to their AI assistant and ask questions of their data, safely scoped to what they're already allowed to see.

✨ What's new
Members can connect the connector and query their data conversationally. To keep this safe, it reads from the tables and charts in your Spaces: when someone asks a question, the AI looks for the answer in the Spaces they already have access to. People can only ever query data that's been shared with them.

⚙️ How it works
We're adding new tools that fetch data from the tables and charts in your Spaces, giving members access for the first time. The existing tools that fetch data directly from your models stay available for admins and super admins.

The new tools also change how data is fetched. Instead of loading an entire table into the conversation, the data is uploaded to a fast query engine and the AI runs targeted queries against it, receiving only the results it asked for. Far less data moves back and forth, which means quicker answers and far fewer tokens, so lower cost and less chance of hitting context limits on longer conversations.

📌 Good to know
MCP access is intended for up to ~20 users per organization. Connecting more works today, but we may introduce limits later — reach out to your account manager if you expect heavy usage.

Planning across a Chart of Accounts or any multi-level dimension no longer means clicking your way back through the hierarchy. Pick the item you want, wherever it sits, and Abacum completes the rest of the path for you.

• Pick the lowest level (the leaf) and you're done. The whole path above completes automatically from your dimension mapping.
• Pick a middle level to finish everything above it, then choose the levels below.

A few things it handles for you:

• Hidden levels: even when the table doesn't show every level, pick the item you need and Abacum fills in the hidden levels above it automatically.
• New combinations: if the item combination you need isn't mapped in your dimension yet, you can create it on the spot, without leaving the table.
• Respects your Space filters: the selector only offers items allowed by the Space, so you stay within the right scope and nothing gets allocated where it shouldn't.

See it in action: Edit dimension items in the Input Table

✨ What’s new

• Rank by top or bottom N. Choose any dimension (clients, vendors, cost centers, regions...), set a direction (top or bottom) and a number. The table shows only those rows, ranked by their contribution to each variable.
• Pick exactly which column drives the ranking. Select the specific time period and version column you want the ranking based on.
• Keep totals accurate with the "Other" group. Items outside your Top N are automatically consolidated into a single "Other" row, so the table total always adds up. You can choose to show or hide it.
• Layer with existing filters. Top N works alongside dimension item filters. If a space-level filter is active, Top N ranks within those results, not across all data.

⚙️ How it works

1. Open the table builder and click + Add filters.
2. Select a dimension, then choose "is in top" or "is in bottom" instead of "is".
3. Configure the rank number, the column to rank by, and whether to show the "Other" group. Click Apply.

👉 Learn how to configure Top N filters in the Academy.

You can now trace where each value in a Dimension came from, directly on the dimension page.

A new Dataset sources column shows, for every value, which dataset (or datasets) introduced it. Before you clean up or debug a Dimension, you can always tell where a value originated.

• When a value is fed by more than one dataset, all of them appear on the same row.
• If an unexpected or unassigned value shows up, you can see at a glance which dataset brought it in.
• Filter the Dataset sources column by one or more datasets to focus on a single source, or select Other sources to surface values that weren't introduced by any dataset.

Learn how to trace dimension value sources

The Public API already returns what your models calculate. Now it also returns how they calculate it: the formula, dimensions, and aggregation behind every variable.

✨ What's new

A new endpoint returns a model's full logic, one entry per variable, the same logic you get from the model download, now available programmatically. Pass a scenario to retrieve that version's formulas.

📈 What you can do

• Access your model logic programmatically to audit how your models are built, for example spotting variables that are missing dimensions, or to document and store it in your own tools.
• Coming soon: access to model logic through the MCP, so Abacum's AI can reason about how your models are built, not just the numbers they return.

⚙️ How it works

Request a model's logic by its id to get every variable with its formula, dimensions, aggregation, and formatting. Add an optional scenario to include that version's formulas. Same OAuth2 authentication, read scope, and super admin access as the rest of the Public API. You can download the full API documentation from Company Settings > API > Download API documentation.

Until today the Public API only gave you model-related information. We are now adding dataset-related data it gives you the data behind them too, both as it first arrives in Abacum and after Abacum has cleaned and enriched it.

✨ What's new

You can now pull dataset data straight from the Public API at two stages. The raw version is the data exactly as it came in from your integration, untouched. The processed version is the same data after Abacum has been transformed, adjusted, and mapped to dimensions, which is the version your models actually use. A new endpoint also lists every dataset in your organization along with when it last synced, so you can pull what you need automatically.

📈 What this unlocks

• Keep your systems in sync: check that the data in Abacum matches your source system, and use Abacum as your single enriched source for everything downstream.
• Feed other tools directly: send clean, ready-to-use data into your BI, analytics, or data science stack without building a model first. You can now use the power of Abacum's ETL to pipe clean and supervised data to your LLM's directly.
• See the full picture: compare the raw and processed versions to understand exactly what Abacum changed.

⚙️ How it works

List your datasets to find the one you want, then request either the raw or the processed version. Abacum returns a secure download link that you can limit to a specific date range. Access works the same way as the rest of the Public API and is available to super admins. The updated API documentation is downloadable from Company Settings > API > Download API documentation. You can also check the documentation from the Academy here.

You can now reference the current year and the current month as plain numbers in your formulas, alongside a new CurrentYear parameter for year-level conditional logic.Your formulas can now reference the current calendar year and month as numbers, opening up a new class of time-based calculations that were not possible before.

✨ What's new

Three new parameters are now available in Model Builder: CurrentYear, MonthNumber, and YearNumber.

CurrentYear brings year-level conditional logic to your formulas, using the same syntax you already know from CurrentMonth. You can now apply different assumptions based on the calendar year, making it straightforward to model year-specific one-offs, phased transitions, or multi-year plans directly in your formula logic.

MonthNumber and YearNumber take a different approach: rather than being used in conditions, they return the current period as a plain number you can use in arithmetic. This unlocks a range of calculations that depend on where you are in the year or how far along a multi-year plan you are.

⚙️ How it works

• CurrentYear works exactly like CurrentMonth in conditional statements: IF (CurrentYear IS 2024) THEN (Revenue.t-1 * 1.08) ELSE (Revenue.t-1 * 1.12). 
• MonthNumber returns a value from 1 to 12 (January = 1, December = 12), and YearNumber returns the four-digit calendar year, both ready to use in arithmetic formulas: YTD Target = Annual Target / 12 * MonthNumber or Base Salary = Starting Salary * (1.03 ^ (YearNumber - 2024)). All three parameters reflect the natural calendar, independent of fiscal year configuration, and are available in both Actuals and Version formulas.