For agencies: projects, runs and credits
Manage clients in one account: configure projects, run monitoring and individual prompts, plan a shared budget and prepare client reports.
One agency account, a separate project for each client
A client project is called a brand in GEO Scout. Each brand has its own prompts, engines, competitors, results and settings. Agency projects share an account credit balance.
- The GEO Scout team enables agency accounts. If these capabilities are missing, contact support@geoscout.pro.
- Open the brand switcher in the top left and choose “Add brand”. Complete the setup: website, business description, geography and prompts.
- Create a separate brand for each client or independent project. Agency accounts have no limit on the number of brands.
- Choose another client in the same switcher. Check the active brand before running prompts, changing settings or exporting data.
- Configure each project’s website, name variants, geography and competitors separately. A shared balance does not combine client results.
Screenshots show a dedicated training account with GEO Scout and Ozon projects. Settings and balances are examples; monitoring results require real runs.
How to add a new client brand
Create a brand from the project switcher while staying in your agency account. The new client’s prompts, results and settings remain separate from other projects.
- Open the current brand name in the top left and choose “Add brand”. This opens the new project setup wizard.
- Enter the client’s website. After analysis, review the business name and description: what it offers, its audience and service area. If the website cannot be analyzed, use the available manual description option.
- Confirm the business type and geography: country or city. Local service recommendations depend on the user’s location.
- Review suggested groups and prompts. Keep questions the client wants to be discovered for, correct wording and remove irrelevant suggestions. Extend the set later on the “Prompts” page.
- Complete setup and check that the switcher shows the new brand. Open “Prompts”, review active questions and choose monitoring engines.
- Agency brands do not automatically receive a free initial run. Review the shared balance, choose “Run all”, check the quoted cost and confirm the first paid measurement. Enable scheduling separately after reviewing settings.
- If colleagues work on the client, open “Settings → Team” and grant access to this brand. For the next client, choose “Add brand” again instead of renaming an existing project.
Before the first run, check the client, geography, active prompts and cost. Adding a brand does not give it a separate balance: its operations use agency credits.
Organize your prompts into groups
Open “Prompts” in the relevant project. Groups separate business topics and customer choices; switches control which prompts participate in regular monitoring.
- Choose “Add group” and name a topic, such as “Choosing a service” or “Comparing solutions”. Add questions using the input within the group.
- Use “In bulk” to add several prompts, one per line. Review the wording and group assignment afterwards.
- Write questions a real customer would ask. Use unbranded questions to measure brand discovery and a separate group to test knowledge of a named brand.
- Check the group’s geography and prompt language. The same question can produce different recommendations for different cities or countries.
- Disabled prompts are excluded from full project runs. Disable one prompt or a whole group while keeping it available for later use.
Agree on a small set of key scenarios with the client first. Expand after reviewing responses: more prompts and engines consume more credits.
Run full project monitoring
“Run all” executes the current brand’s active prompts and collects data for project analytics. A manual run does not enable a recurring schedule.
- Select AI platforms under “Monitoring engines”. Claude is marked as costing 2 credits; ordinary engines cost 1 credit per prompt.
- Check active prompts and selected engines. If rotation is configured, the full run uses the current slot’s engines.
- Choose “Run all”. Review the scope, cost, available balance and remaining balance in the confirmation dialog.
- Confirm and wait for completion. Another full run is blocked while the project is analyzing; response times vary between platforms.
- Find results under “Responses” and in project analytics: Overview, Competitive intelligence, Brand perception and Citations. Check the date range and filters if new responses are missing.
Example: 6 active prompts × 3 ordinary engines = 18 credits per full run. Adding Claude increases the cost to 30 credits: 6 × (1 + 1 + 1 + 2).
Test one prompt or a cluster in the sandbox
The sandbox is for quick wording and topic checks. Unlike full project monitoring, each targeted run has a separate entry in sandbox history.
- Choose “Run” next to a saved prompt. To test a group’s active prompts, choose “Run cluster”.
- These actions use the brand’s selected engines and open the sandbox. A cluster creates one run per prompt; if only part starts, check how many prompts actually launched.
- On the “Sandbox” tab, select a saved prompt or enter a new question and choose engines for the test. A new question is saved as a project prompt: review its group and active status after testing.
- Review the cost before running: sandbox checks use the same agency balance. Repeating a question also consumes credits.
- Review response text and run history in the sandbox. For ongoing tracking and project analytics, use saved active prompts with “Run all” or a schedule.
Test a new question on the relevant engines in the sandbox before adding it to the permanent monitoring set.
Set a schedule for each client
The schedule switch sits above the prompt list, next to the manual run button. Scheduled runs use the current project’s active prompts and engine settings.
- Finish configuring prompts and make an initial test run, then enable the project schedule.
- Agency accounts can choose daily or weekly runs. Check the time shown in the schedule tooltip; the Russian interface displays Moscow time.
- Set frequency separately for each brand. For example, check a primary project daily and a pilot project weekly.
- Disabling a schedule stops future automatic runs for that project. It does not cancel a run already started or disable other brands.
- The toolbar shows estimated spending and how long credits should last. Recalculate the budget after changing prompts, engines or frequency.
Schedules are disabled in the training account shown here. Enable schedules on working projects after checking configuration and budget: automatic runs consume the shared balance.
How agency credits work
Credits pay for operations in an agency account. The agency owns the balance, and its projects draw from it without buying a separate package for each client.
- One prompt run on an ordinary engine costs 1 credit; Claude costs 2. A full run costs the active prompt count multiplied by the sum of selected engine costs.
- Manual monitoring, scheduled runs and sandbox tests consume credits. Additional charges include report generation at 10 credits, fact checking at 1 credit per answer, and knowledge base indexing at 1 credit per source.
- “Billing” shows the shared balance, used credits and “Buy credits”. This balance covers all agency projects.
- The current price is ₽5 per credit. Select a preset package or enter between 100 and 50,000 credits; review the total before paying.
- Pay by card, or by invoice with business details for a legal entity or sole proprietor. Credits are added after payment is confirmed, rather than when a checkout or invoice is opened.
- If the balance is insufficient, reduce the run scope or top up. Failed monitoring and sandbox requests have technical failure refunds; check the final outcome in transaction history.
The shared agency balance differs from credits on a regular brand subscription. Use the account balance and the quoted operation cost when planning agency work.
Plan the budget and track spending by client
Spending depends on prompt count, engine prices and run frequency. Calculate one run first, then add all planned runs across projects.
- Example: 20 prompts × 3 ordinary engines = 60 credits per run. Daily monitoring over 30 days costs 1,800 credits, or ₽9,000 at the current credit price.
- Four runs of the same set cost 240 credits, or ₽1,200. For a weekly schedule, count the actual run dates in the month.
- Add all project budgets and allow extra for sandbox tests, manual runs and reports. Two projects costing 60 credits per run together consume 120.
- Open credit history in “Billing” to see the date, brand, amount, operation type and description. Top-ups without a brand belong to the whole account.
- Filter by brand and export history to reconcile a client’s spending. Compare planned spending with actual charges and refunds, as well as the forecast.
The balance forecast is approximate: one-off operations and configuration changes affect it. Keep enough reserve for several runs, particularly when projects run on the same day.
Give your team and clients access to selected projects
Open “Settings → Team”. Invite someone to one or several brands without sharing the agency account password.
- Choose “Invite member”, enter an email and select the projects. Selecting several brands sends one invitation.
- Review the selection before sending: give clients their projects and colleagues the projects they work on.
- Members can access data and monitoring settings. This is working access rather than a read-only role: consider their ability to change settings and start paid operations.
- Team management and brand deletion remain with the owner. After an invitation is accepted, change the brand selection using “Access”, or revoke access.
- Operations in an agency project consume its owner’s balance. Agree on run policies with colleagues to avoid unnecessary duplicate checks.
Use an export or report when someone only needs the results. A project invitation is for collaboration and provides broader access than reading a PDF.
Prepare a client report
After collecting responses, generate a project report using “Report” next to “Run all”. Generation costs 10 credits.
- Select the brand and choose “Report”. Pick a period that already contains monitoring results; a window without responses cannot be used.
- Enable comparison to select an earlier period with data. The comparison period must end before the main period starts.
- Review dates, cost and balance, then confirm generation. The report is prepared in the background.
- Find completed documents under the project Overview’s “Reports” tab. Check the brand, dates, prompt coverage and conclusions before sharing.
- For agency styling, open “Report branding” in the user menu and configure the available branding options before generating the report.
Keep prompts and engines comparable between periods. Explain any monitoring scope changes to the client because they affect conclusions about trends.
An agency workflow from setup to the next measurement
Use the same sequence for every client to control configuration, spending and the connection between completed work and AI visibility changes.
- Setup: create the project, review its description and geography, and agree on customer questions and competitors.
- Baseline: select engines, review the cost, run full monitoring and inspect responses, mentions and citations.
- Action plan: identify visibility and perception gaps, select Command Center tasks and prepare website improvements or publications.
- Tracking: enable an appropriate schedule, test new ideas in the sandbox and reconcile spending against credit history.
- Reporting: collect the next comparable measurement, prepare a report and agree on further actions with the client.
Before every operation, check the active project, run scope and cost. This becomes especially helpful when managing several clients in one account.