> ## Documentation Index
> Fetch the complete documentation index at: https://clearskies.cc/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Quick Start

> Connect your data, build your graph, and put it to work in about 10 minutes

Everything in Clearskies starts the same way: connect your data sources and your Context Graph builds. What you do next depends on how you want to work. This guide covers setup, then each path.

## Before you begin

You'll need:

* Admin access to your data sources (Salesforce or HubSpot, plus Gong or Google Workspace, are good places to start). Not an admin?
* About 10 minutes

## Set up: Connect your data sources

Connect your CRM plus at least one conversation source, calls or email. Your CRM knows the deal exists. Your calls and email know how it's actually going. The graph connects the two.

1. Log in to Clearskies (app.clearskies.cc)
2. On **Home**, go to **Data sources**
3. Connect **Salesforce** or **HubSpot**
4. Connect a conversation source: **Gong**, **Scratchpad**, or **Fathom** for calls; **Google Workspace** or **Microsoft 365** for email and calendar

**What's happening:** Your Context Graph is building. Watch the **Context graph** panel on Home fill in with accounts, people, deals, meetings, emails, and calls, and the relationships between them. Clearskies resolves identities across sources, so the contact in your CRM, the voice on the call, and the name on the email thread become one person. Each source shows **Live** once it's syncing. Most sources sync quickly; sources with a large call history, like Gong, take longer to backfill.

Once your sources are Live, pick your path.

## Path 1: Chat with your graph

In Claude, ChatGPT, or whatever your team runs, your AI answers with every account, deal, and call behind it.

1. Connect over [MCP](/docs/building-with-clearskies/mcp-server): Claude (Claude.ai or Claude Desktop) or ChatGPT (developer mode)
2. Ask something you'd normally have to open your CRM to answer: a deal you're working, an account you own, what's on the calendar this week

Answers cite their sources, pulled from every system you've connected. For getting the most out of your graph, see best practices.

## Path 2: Automate your workflows

A brief before every call, a summary after, without anyone asking. Workflows run on triggers (a meeting starting or ending, a schedule, a Salesforce record changing) and deliver to Slack, email, or your CRM.

1. Build and publish an **agent**, the reasoning step your workflow will call
2. Create a **workflow**: pick a trigger, add your agent, choose where the result goes (Slack, email, Salesforce, HubSpot)
3. Test it (test runs are always dry, nothing sends or writes), then publish

Start from a pattern your team feels weekly: a pre-call brief, a pipeline risk digest, QBR prep. More in what you can build.

## Path 3: Run an analysis

Questions that span your whole book of business, not one deal, answered from your real data, every claim cited.

From your connected AI, run:

* **Win-loss analysis**: why you win and why you lose, across every closed deal
* **Competitive intelligence**: which competitors show up in your calls and how deals against them go
* **Rep coaching**: patterns across a rep's real conversations, with timestamps

## Keep going

**Connect the rest of your stack.** Slack conversations, Pylon support tickets, Jira and Linear engineering work. Every source makes every answer, agent, and workflow more complete. → Context Graph

**Build in your own tools.** The same MCP connection works anywhere MCP does: automations in n8n, internal tools and dashboards in Retool, anything else that speaks MCP.

## Need help?

* Visit [Getting Help](/docs/support/getting-help) for troubleshooting and support