The Factors MCP (Model Context Protocol) server lets you bring your Factors data and agents into Claude. Once connected, you can ask about your accounts, journeys, attribution, and engagement signals in plain language, and Claude pulls the answers live from Factors for a particular account. You can also pull your saved reports, funnels, events, and KPIs directly into chat and have the LLM analyze them for you. This works across MCP-compatible clients, and works best in agentic coding clients like Claude Code or Codex.
You can setup MCP in two ways, via custom connector or via package.
Setting up MCP via Custom Connector
This guide uses Claude as the example, but the same MCP URL works with any MCP-compatible client.
What you'll need
A Factors account with access to AI Features
Your Factors personal access token (Settings, then AI Features)
The Factors MCP server URL:
https://mcp.factors.ai/mcpA Claude account
Step 1: Copy your personal access token
Log in to Factors.
Go to Settings, then AI Features.
Copy your personal access token.
Step 2: Add the Factors custom connector in Claude
Open Claude and go to Settings, then Connectors.
Click the plus button next to Connectors and choose Add custom connector.
Enter the details:
Name: Factors
Remote MCP server URL:
https://mcp.factors.ai/mcp
Click Add.
On Team and Enterprise plans, an Owner adds the connector once under Organization settings, then Connectors, and each member connects with their own token.
Step 3: Connect
In your Connectors list, find Factors and click Connect.
When prompted, authenticate with your personal access token.
Factors will show as connected, with its available tools listed under the connector.
Step 4: Use Factors in a conversation
Start a new chat in Claude.
Make sure the Factors connector is enabled for the conversation.
Ask questions using your Factors data
Connect your LLM to Factors and ask questions in plain language. The Factors MCP sends each question to the right capability and returns the data. Your LLM then does the analysis.
What you can ask
Capability | Use it for | Example |
KPIs, funnels and traffic | Counts, trends, conversion rates, website sessions, ad spend and clicks | How many MQLs did we get this month? |
Influence attribution | Which channels, campaigns or keywords touched pipeline, revenue or deals | Which LinkedIn campaigns influenced pipeline this year? |
Saved reports | Running reports your team has already built in Factors | Run the “Top accounts by engagement” report for last 30 days. |
Agents | Researching a specific account by its domain | Run the account research agent on acme.com. |
Tips for better answers
Name the metric you want: MQLs, pipeline, spend, sessions, and so on.
Give a time range. If you leave it out, most questions use the last 7 days.
Name the specific channel, campaign, account domain or report.
Use exact report names when you want to run a saved report.
KPIs, funnels and traffic
Use this when you’re asking “how much”, “how many” or “what rate”.
Supported
KPI counts and trends, such as MQLs, SQLs, pipeline and deals closed
Funnel conversion and drop-off between events
Website sessions, visitors and event activity
Ad platform metrics: spend, clicks, impressions, CTR and CPL
Filters and breakdowns, for example by campaign, source or country
Any time range, plus period comparisons (WoW, MoM, QoQ, YoY)
Questions with several parts, answered from one prompt
Good to know
If your wording could match more than one KPI or event, you’ll be asked to choose the one you mean before the report runs.
Example prompts
How many MQLs did we get this month, compared with last month?
Conversion from website session to form submission over the last 30 days, broken down by campaign.
What did we spend on LinkedIn last quarter, and what were the CTR and CPL?
Web visitors last week by source.
Influence attribution
Use this to see which channels, campaigns, ad groups or keywords touched your pipeline, revenue, leads, MQLs, SQLs or closed-won deals. You can also ask about ROI, cost per conversion, and how to split budget between LinkedIn and Google.
How credit works
Factors uses influence attribution, which gives credit to every touchpoint in the journey. These models are not supported: first touch, last touch, linear, time decay, U-shaped, W-shaped and custom.
Supported options
Option | Details |
Channels | Paid Search, Paid Social, Paid Video, Paid Display, LinkedIn (all by default) |
Level of detail | Channel, campaign (default), ad group, keyword (Paid Search only) |
Time range | One calendar year, 2025 or later: this year, last year, or a named year |
Metrics | One conversion metric per question |
Number of rows | Top 5 by default (top 3 for channels), up to 20 |
Sorting | By conversions (default), spend, clicks, impressions or CPC. Words like “bottom” or “worst” sort from the lowest. |
Good to know
If you ask for a quarter, a month range or year-to-date, you’ll be asked to pick this year or last year.
If you ask for two metrics at once, such as pipeline and revenue, you’ll be asked to choose one.
If you ask for “last 7 days” or “this month”, you get results for the current year. Results can’t be split by week or month.
To look at one campaign, put its name in quotes, for example campaign “Competitor Terms”.
Example prompts
Which LinkedIn campaigns influenced pipeline this year?
Cost per SQL for each LinkedIn campaign this year.
Which Paid Search keywords influenced closed-won deals last year?
Compare cost per pipeline dollar for LinkedIn and Google this year.
Bottom 5 campaigns by revenue influence this year.
Saved reports
Bring the saved reports from your Factors project straight into the chat. You don’t need to open Factors.
How it works
Ask about a report, funnel or KPI.
Your LLM finds the matching saved report by its title.
The report runs for the date range you gave and the data comes back to the chat.
Your LLM summarizes or analyzes the results.
You can run one report for a quick readout, or several at once and get a combined analysis. For example, “Give me spend, pipeline influence and view-through conversion for LinkedIn” pulls each matching report and brings the results together.
Getting the right report
Reports are matched by title, so using the actual report name gives the most reliable results.
Works well | Less reliable |
Run the LinkedIn ad viewed → Website session → Form submitted report | Show me how LinkedIn ads convert |
If more than one report could match, your LLM should check with you before it runs one.
Good to know
Access: you can see every saved report in the project, including reports other people built.
Read-only: you can’t create, edit or delete reports from the chat. Use the Factors app for that.
Data: Factors returns the raw report data, and your LLM does the analysis.
Date range: if you don’t give one, it defaults to the last 7 days. Long ranges can time out. If a run fails, try a shorter window.
Example prompts
Run the LinkedIn ad viewed → Website session → Form submitted report for the last 30 days.
Pull my “Top accounts by engagement” report and summarize the top 5.
Show me last month’s SQL funnel report and tell me where the biggest drop-off is.
Give me spend, pipeline influence and view-through conversion for LinkedIn.
Account research with agents
Run your published Factors agents against any account domain. Depending on the tools an agent was built with, it can summarize the account’s journey, do web research or find contacts.
What you can do
List the published agents in your project
Run an agent on a company domain, such as acme.com
Create a new agent or update an existing one from the chat
Example prompts
What agents do we have?
Run the account research agent on acme.com.
Summarize the account journey for acme.com using our journey agent.
Create an agent that summarizes an account’s journey over the last 30 days and pulls recent news about the company.
Troubleshooting
Connection fails: confirm the URL is exactly https://mcp.factors.ai/mcp and that you used the full token from Settings, then AI Features.
Claude isn't using Factors: open the tools menu in the chat and confirm the Factors connector is toggled on.
Empty or off results: make the prompt more specific with names, a time range, and the exact metric.
Installing Python 3.11 or later and configuring the Factors MCP server across Mac, Windows, and Linux
Prerequisites
Before starting, you will need:
A Factors Agents Authorization Header
Terminal / Command Prompt access
AI application - here we will go with Claude Desktop as an example for setting up MCP.