Jepto MCP Server Overview

Jepto exposes a remote MCP (Model Context Protocol) server so an external AI assistant can call tools against your Jepto account: list clients, query connected marketing data, read KPI and budget status, search client knowledge, load guidance on how each data source behaves, search this Help Center, build and edit reports, and (with confirmation) create or update KPIs and budgets and run platform actions.

The server uses the same Data Model field names as Reports and Data Chat (camelCase fully qualified names, e.g. googleAds.spend). See Data Model and Data Query for how those queries are shaped in the product.

Not the same as Data Chat MCP connectors. This article is Jepto's account MCP server (https://mcp.jepto.com). Data Chat can attach separate MCP servers (Google Sheets, Docs, etc.) per conversation; those connectors are documented under Data Chat in the help center, not on this page.

Why use the Jepto MCP

• One connection for every brand you manage. The connection is account-wide, so a single question can compare KPIs, budgets and performance across your whole portfolio without connecting each brand separately.
• The same data as Reports and Data Chat. Metrics use one definition across platforms, so an answer in your AI assistant lines up with what you see in Jepto.
• Budgets and KPIs are tools, not just data. The assistant can list, create and update budget targets and KPIs, with a preview before anything is saved.
• Changes need your confirmation. KPI, budget and platform action changes show a preview first and only run after you confirm.
• Reports from a conversation. The assistant can build and edit Jepto reports: tabs, charts, tables, layout, theme and analysis. It can also show the finished report right in the chat.
• Built-in guidance. Skills tell the assistant how each data source behaves, including what it cannot report, and the connection flags sources that are still syncing so partial data is called out.
• More than ad platforms. Where connected, your own BigQuery data, Xero, Google Business Profile and AI Visibility results can be queried through the same tools.

Connection

Setup guides

• Connect Jepto MCP to Claude: add Jepto from the Claude connector directory and sign in to Jepto.
• Connect Jepto MCP to ChatGPT: install the Jepto plugin from ChatGPT's Plugins page and sign in to Jepto.

Other products that support custom remote MCP connectors may work with the same URL; steps depend on the client. We document Claude and ChatGPT step by step today.

How the assistant uses it

1. Your MCP client discovers 44 tools exposed by mcp.jepto.com after connection.
2. You ask a question in natural language; the model selects one or more tools (you do not need to name tools yourself).
3. Read-only tools return JSON (account, client, analytics rows, KPI/budget lists, knowledge search hits, skills, help articles).
4. Write tools (kpi_forecasting, budget_tracker, actions_execute) first return a preview; you confirm or cancel in the client's MCP App UI. Pending confirmations expire after about 10 minutes. Nothing is applied to Google Ads (or saved in Jepto) until you confirm.

For analytics, the usual discovery order is:

client_list → datasource_list → analytics_schema → analytics_query

Use client_list to obtain a 16-character hexadecimal clientId before any client-scoped call. Do not invent IDs. For a source with known quirks, skills_get returns the guidance (for example googleAds, facebookAds or analyticsQuery) before the query is written.

Available tools

Tool names below are the exact MCP names your client will show.

Account and clients (read-only)

Data sources and analytics (read-only)

datasource_list service values (e.g. googleAds) align with sources on analytics tools. dataSourceId is for KPI and budget tools, not analytics queries.

Reports (read + write)

The report tools let the assistant find, build and change Jepto reports. It can create a report and its tabs, add chart, table, text, image and analysis cards, lay them out, apply your theme and write the analysis. Reports made this way are normal Jepto reports: you can open them in the editor, change them with Report Chat and share them. In clients that support MCP Apps, view_report shows a report right in the chat, so you can check it without leaving the conversation.

Report changes are saved straight away, without the preview step that KPI and budget changes use. Page, card and report updates check the latest saved version. If someone else saved a change first, the assistant gets a conflict, re-reads and tries again, so it doesn't overwrite their work.

A new report is built in order: report_create makes the tabs, analytics_query checks each metric has data, report_card_generate adds the cards, report_page_update places them and report_analysis_run writes the summary. view_report then shows the finished report in the chat. Chart types are number, line, area, bar, column, donut and scatter. Map cards can't be created over MCP yet. For example prompts and tips, see Build Reports with Jepto MCP.

KPIs and budgets

kpi_forecasting is a single tool with read actions (list, info, read) and write actions (create, update, delete), not two separate tools.

Platform actions (preview + confirm)

Actions for other platforms, and changes to budgets inside an ad platform, are not available today.

Skills and Help Center (read-only)

Web (read-only)

Permissions and safety

• Tools only access data and connections already in your Jepto account.
• Connection is scoped to the account, not a single client: without a clientId, list tools can return all clients, KPIs, or budgets the login can see.
• KPI, budget and platform action changes go through confirmation UI, not silent execution. Report changes are saved straight away.
• Platform credentials (Google, Meta and others) stay in Jepto and are never sent to the AI assistant. Access tokens last 60 minutes and managed clients refresh them automatically. See OAuth & Authentication.
• In Claude, you can choose whether the model must ask before each tool call; many teams use always allow only after they are comfortable with the tool set.

Limits to know about

• Usage: Jepto MCP has no usage limits on any plan.
• Platform actions: limited to enabling and pausing Google Ads campaigns.
• Sign-in: OAuth with PKCE only. Clients that cannot complete that flow cannot connect, and there is no API key option.
• Speed: analytics run against live data. Cross-source queries (such as unified.* metrics across many connected sources) can take a minute or more, especially on first run.
• Source coverage: some sources expose fewer fields than others (X Ads, for example, has only a few metrics and dimensions). The matching skill lists what is missing.
• Setup guides: step-by-step instructions exist for Claude and ChatGPT. Other clients that support custom remote MCP connectors may work with the same URL.

Use cases

Morning portfolio check

Ask which KPIs or budgets are in critical or warning status across the account. The model typically calls kpi_forecasting and budget_tracker with action=list and status filters, then summarizes by client.

Client call prep

Before a meeting, ask for a client overview plus recent performance. Expect client_info, optional client_knowledgebase searches, and analytics_query for key metrics over a defined range (e.g. last 30 days vs prior period).

Ad hoc analysis without building a report

When you need a one-off breakdown (campaign spend, channel mix, Search Console trends), use conversational follow-ups. The model discovers fields with analytics_schema, then runs analytics_query with filters and sort, using the same data model as chart cards, without publishing a report.

Build a client report from chat

Ask for a report for one brand, the tabs you want and a date range. The model calls report_create, checks each metric with analytics_query, adds cards with report_card_generate, places them with report_page_update and can finish with report_analysis_run. You get an editor link and a preview link, and view_report can show the report right in the chat. See Build Reports with Jepto MCP.

Cross-channel spend check

Ask for total spend and clicks across every connected ad platform for one brand and period. The model queries the unified metrics, which roll spend, clicks, impressions and conversions up across platforms. These queries can take longer than single-source ones.

Create or adjust a KPI from chat

Describe the metric and target; the assistant resolves metricName from schema and datasource rows from datasource_list, then calls kpi_forecasting with action=create. You review the preview and confirm in the MCP App.

Pause or enable Google Ads campaigns

After analysis in the same thread, the model can call actions_list, then actions_execute with googleAdsPauseCampaign or googleAdsEnableCampaign and a reason summarizing the conversation. You confirm the exact campaigns in the UI before Google Ads changes.

Find institutional context

Search what Jepto already stores about a client (brand guidelines, target audience, QBR notes) via client_knowledgebase instead of hunting through emails or docs.

Getting started

1. Connect https://mcp.jepto.com in your MCP client and sign in to Jepto (Claude steps, ChatGPT steps).
2. Ensure clients and data sources are connected in Jepto. MCP analytics tools only see what the account already has.
3. Ask a concrete question (client name or list my clients, date range, platform, metric). Use follow-ups to narrow breakdowns or change the period.

No special command syntax is required; clarity about client, platform, and time range improves results.

Last verified against the live server on 2 November 2026 (44 tools).