Getting the most out of Gatlas

An assistant knows nothing about your companies beyond what Gatlas tells it. Open each company on Entities and fill in Agent instructions under AI agent context. Assistants read it when they start working with your data. Write what a new analyst would need on day one:

• What the company does and how it makes money.
• Its fiscal year, reporting currency and accounting system.
• Anything that trips people up: a second legal entity, a one-off write-down, a customer who pays annually in advance.

Admins and above can edit it. See the onboarding guide for where it sits.

On a view, AI context is the single field that changes answers most: every question and every skill that touches the view starts with that text. Say what one row is, what each figure means, which currency it is in, and what the view does not include. "Revenue excludes VAT and refunds; one row per company per month; amounts in the company's own currency" prevents more wrong answers than any prompt. See Views and SQL.

AI assistants only see published views. MCP is how AI assistants such as Claude connect to Gatlas, so the switch is Publish to MCP when you create a view and Published to MCP when you edit one. Use it as a quality gate, not a default:

• Publish a view once you have checked one of its numbers against the books.
• Keep one view per question. Two views that both claim to be "revenue" give two answers.
• Name views plainly, with the company first: "Northwind Monthly P&L", "Northwind Invoices". The skills find their views by name.
• Unpublish or delete views you have replaced.

Four habits make most answers better:

• Name the company. "Northwind", not "the retail one".
• Name the period. "Q2 2026" or "the last 12 months". Open-ended dates get vague answers.
• Name the view when you know it. "Use the Monthly P&L view" keeps the assistant on your curated numbers.
• Define your terms. "Overdue", "active customer" and "margin" mean different things in different places. Say which you mean, or put the definition in the view's AI context once.

Prompts that work well to start from:

Which views can you see, and what does each one answer?

What was revenue by company for each of the last six months, and which company grew fastest?

List Northwind's invoices more than 90 days past due, largest first, with the customer and the amount.

Compare Northwind's gross margin this quarter with the same quarter last year, and explain the biggest change.

Is our data current? Flag any company whose sources have not synced recently or whose books have gone quiet.

If an answer looks wrong, ask the assistant which view and filters it used. That usually shows the problem in one step. The setup for each assistant is in Using your MCP connection.

• A green sync is not current books. It means Gatlas reached the tool. For a company with a ledger view, if nothing has been posted in its accounting for more than 30 days, Gatlas warns the assistant, and you should treat "this month" answers for that company with care.
• Adding companies up is not a consolidation. A total across companies counts both sides of any trade between them. Gatlas does not eliminate those for you.
• Currencies need a rule. Each company reports in its own currency unless a view converts it. Before comparing companies, check which rate the view uses.
• Long lists come in pages. An assistant reads up to 500 rows of a view at a time and is told when there are more. For totals, ask it to calculate the total rather than add up a list.

A skill is a finance workflow, such as a monthly P&L, an aging report or a portfolio rollup, written out step by step so Claude follows the same method every time. Copy one from Skills and paste it into Claude. See Skills.

If you ask for the same charts every month, ask Claude to build them once as an artifact: a dashboard, an A4 report or a deck that reads your views each time it opens. Share it with a link or print it. See Artifacts.

Numbers explain what happened; board notes, memos and KPI definitions explain why. Sync a Google Drive folder per company and distill it into short documents, so an assistant can answer "why did margin fall?" with the reason as well as the figure. See Documents.

Connect a Slack channel per company. Alerts post when a sync fails or a connection goes stale, a weekly digest sums up the week, and a scheduled artifact can land in the channel on the first of every month. See Notifications and reports.

Access drifts as people join and leave. Once a quarter:

• Under Settings, Members: check each person's role and companies, and remove anyone who has left. See Team and access.
• On MCP: sort by Last used, and delete connections nobody uses with Delete connection. Check that anyone who only asks questions has a Read-only connection. See MCP connections.
• On Sources, under Setup links: revoke links that are still pending and no longer needed. See Connecting sources.
• On each artifact you have shared: make sure the embed link still belongs where it was sent.

Activity, reached from View all activity on Entities, is the log of the changes you can see, if you need to see who did what.

Use Feedback in the app's sidebar, or write to team@gatlas.ai; we usually reply within a business day. You can also ask Claude to send us feedback from inside a conversation, which reaches the same people.