Connect your AI assistant to Clarisights

Bring your Clarisights data into the AI tools you already use. Once connected, you can ask Claude, ChatGPT, Cursor, or any MCP-compatible client questions about your campaigns, metrics, and reports, and get answers grounded in your live data.

Clarisights exposes a remote Model Context Protocol (MCP) server. If your AI tool supports custom remote connectors, you can plug it in.

Connector URL


https://app.clarisights.com/mcp

Paste this URL into your AI tool's Add custom connector flow, sign in with your Clarisights account when prompted, and you're done.

What you can ask

The server gives your AI assistant read-only tools. Once connected, it can browse the same data you can: your data sources, reports, dashboards, metrics, dimensions, and saved segments. A few examples to try:

Performance & pacing

What was my Facebook spend yesterday vs. the same day last week?
Show me top 10 Google Ads campaigns by ROAS in March, broken down by week.
Which campaigns went over their daily budget last weekend?

Cross-channel analysis

Compare CPM across Facebook, TikTok, and Snapchat for the last 30 days.
What's my blended CAC across all paid channels this quarter?

Reports & dashboards

List the dashboards I have access to.
Open the Q1 Acquisition Review dashboard and summarize what each tab is tracking.

Discovery

What metrics do I have available for LinkedIn Ads?
Which segments has my company set up?

The assistant finds out which data sources, metrics, and dimensions are available by calling Clarisights directly, so you don't need to memorize field names.

Authentication

🚧 Access is currently gated. To enable access for your organization's workspace, reach out to your CSM.

Clarisights uses OAuth 2.0. When you add the connector, your AI tool opens a browser window where you sign in with your Clarisights account and approve access. There are no API keys or tokens to manage.

Your AI tool stays signed in and renews its session automatically until you remove the connector (see Manage access).

What the AI can see

The MCP connector respects your Clarisights permissions. Your AI assistant can only access data that you can already see in the Clarisights UI:

Data sources and accounts assigned to you
Reports and dashboards shared with you
Segments and metrics defined for your company

It cannot see other users' private dashboards, data outside your assigned scope, or anything from other companies. Changes to your access in Clarisights apply to the AI assistant immediately.

Set up your AI tool

Pick the tool you use. Each setup takes under a minute.

🎥 Watch the setup walkthrough

Claude (web, desktop, and mobile)

Anthropic's Claude supports remote MCP connectors on Pro, Max, Team, and Enterprise plans.

Team and Enterprise workspaces: A workspace admin needs to add the Clarisights connector first, following the same steps below. Once it has been added at the workspace level, individual users can find Clarisights under Settings → Connectors and sign in with their own Clarisights credentials.

Open Claude → Settings → Connectors.
Click Add custom connector.
Enter:
- Name: Clarisights
- Remote MCP server URL: https://app.clarisights.com/mcp
Click Add, then sign in with your Clarisights account in the browser window that opens.
In any chat, click the + button → Connectors → enable Clarisights for that conversation.

ChatGPT

ChatGPT supports custom MCP connectors on Business, Enterprise, Edu, and Pro plans.

Open ChatGPT → Settings → Connectors → Advanced settings → enable Developer mode (if not already on).
Go to Settings → Connectors → Create.
Enter:
- Name: Clarisights
- MCP Server URL: https://app.clarisights.com/mcp
- Authentication: OAuth
Save, then sign in with your Clarisights account.
In a new chat, open the tools menu and toggle Clarisights on.

📝 Note: Custom connector availability depends on your ChatGPT plan and workspace policy. If you don't see the option, ask your workspace admin to enable it.

Cursor

Cursor supports remote MCP servers natively:

Open Cursor → Settings → Cursor Settings → MCP & Integrations.
Click Add new MCP server.

In the JSON editor, add:


{
  "mcpServers": {
    "clarisights": {
      "url": "https://app.clarisights.com/mcp"
    }
  }
}

Save. Cursor opens a browser window for you to sign in with Clarisights.

Claude Code

One command:


claude mcp add --transport http clarisights https://app.clarisights.com/mcp

The first time you use a Clarisights tool, Claude Code prompts you to sign in.

Codex

OpenAI Codex CLI configures MCP servers in ~/.codex/config.toml. Add this entry:


[mcp_servers.clarisights]
url = "https://app.clarisights.com/mcp"

Restart your Codex session. The first call to a Clarisights tool starts the OAuth sign-in flow.

Other clients

Any MCP-compatible client that supports remote HTTP servers works. Point it at https://app.clarisights.com/mcp and use OAuth 2.0.

For older clients that only support local stdio transport, you can bridge with mcp-remote:


{
  "mcpServers": {
    "clarisights": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://app.clarisights.com/mcp"]
    }
  }
}

Tools

Your AI assistant gets these read-only tools:

Tool	What it does
`list_data_sources`	Lists the data sources you can query
`list_metrics`	Finds metrics, with descriptions, most-used first
`list_dimensions`	Finds dimensions and the data sources each applies to
`query_data`	Runs a query and returns the results
`list_reports`	Lists the reports and dashboards you have access to
`get_report`	Reads a report and its tabs

The assistant sees exactly the data the signed-in user can see in Clarisights. Company admins can turn MCP access on or off for individual users in the Manage Users console.

Limits

To keep responses fast and protect your account, the MCP server applies the following limits:

Limit	Value
Requests per minute (per user)	10
Requests per hour (per user)	100
Requests per day (per user)	1,000
Requests per minute (per company)	30
Requests per hour (per company)	300
Requests per day (per company)	3,000
Maximum rows per query	1,000
Query timeout	120 seconds

If a query times out or hits the row cap, the assistant is prompted to narrow the date range, reduce the number of dimensions, or add filters and try again. This keeps it from overloading its context.

Manage access

Disconnect from one AI tool

In your AI tool's connector settings, remove the Clarisights connector.

Revoke access for your whole organization

Reach out to your CSM, or email support@clarisights.com, asking to revoke MCP access for your organization.

Admin controls

Company admins can disable MCP access for individual users from the Manage Users console in Clarisights.

Troubleshooting

Authentication failed or stuck on the sign-in page

Remove the Clarisights connector from your AI tool and add it again, then sign in when prompted. If you connect through mcp-remote, also clear its cached credentials first:


rm -rf ~/.mcp-auth

Tool not available, or your assistant says it can't reach Clarisights

Make sure your AI tool is on a plan that supports custom remote connectors (see the per-tool notes above), that the connector is enabled for the current chat, and that MCP access is enabled for your Clarisights user.

Rate limit exceeded

Wait until the next minute or hour resets. If you hit limits frequently, contact your customer success manager. Limits can be adjusted to fit your use case.

Query times out

Try a smaller date range, fewer dimensions, or add an account filter. Cross-channel queries with many dimensions are the most expensive.

A metric or dimension is missing

The connector exposes the same fields you see in the Clarisights UI. If a metric is missing, check that:

You have access to the relevant data source
The metric exists in your company's setup (custom metrics need to be created in Clarisights first)
For channel-specific metrics, you mentioned the channel in your prompt, e.g. Facebook impressions rather than just impressions

Context management best practices

Your AI assistant reads the metadata you manage in Clarisights, including:

Dimension descriptions
Metric descriptions
Report and widget titles
Text widget content
Custom report segments

If the assistant doesn't pick the right metric or dimension, or keeps going off track, review your metrics and dimensions setup in Clarisights. Add descriptions, especially where several similar-sounding fields exist.

Good

Clear descriptions, e.g. for Country: "Defines the country that the campaign is targeting. Multi-country campaigns have the value set to Multi."
Well-defined metric names, e.g. only one Spend, Impressions, and Clicks custom metric each.
Clear conversion semantics, e.g. for your Purchase metric: "The primary Purchase conversion metric. This counts every purchase made in web or app, attributed to the date of the purchase. This is the last event in our funnel SignUp → Add to Cart → Purchase. Available at Campaign, AdGroup, and Ad levels for most channels. This attribution is provided by our data science team."

Bad

Unclear or missing descriptions, e.g. Country and AF Country with no descriptions. The assistant gets confused and picks one at random, or burns tokens trying all of them.
Unclear metrics with lazy definitions, e.g. Ad Spend, Total Cost, and Media Spend, all described as "Amount Spent".
Purchase v1, Purchase v2, Purchase v3… with no indication of which one is relevant or what each is for.

A rule of thumb for admins: if your users are often confused about which metric or dimension to use, or keep asking which one is right, your AI assistants will likely run into the same issues.

Need help?

Contact us from the ? icon at the bottom right of Clarisights (Send us a message), or email support@clarisights.com. Please include:

The AI tool you're using and its version
The exact error message or a screenshot
The prompt that failed
When the issue started