Scarf AI

Scarf AI helps you ask questions about your Scarf data and use those answers inside the tools where your team already works. You can use Scarf AI directly in Scarf, from Slack, through the Scarf API, or from your preferred LLM using the Scarf Skill or Scarf MCP server.

Use Scarf AI when you want to:

summarize package, company, web, and custom telemetry activity
investigate account or company-level usage
identify high-intent prospects and customer signals
create, reuse, and explain filters for company insights
bring Scarf context into sales, success, developer relations, or security workflows

Runs and billing

Scarf AI uses Scarf Runs. Scarf AI in-app and the Scarf Slack Agent consume 1 Run per message sent to the AI. API calls made by Scarf AI while answering that message do not consume additional Runs, which makes the in-app and Slack AI experiences the most cost-effective way to consume inference with Scarf.

Viewing dashboards, changing filters, or reviewing already-available results does not by itself consume Runs.

For the full billing model, see Billing and Pricing.

Using Scarf AI directly

Scarf AI in-app

Scarf AI is available inside the Scarf application for teams with access to AI features.

To use it:

Sign in to Scarf.
Open your organization.
Open the Scarf AI experience from the app.
Ask a question about your Scarf data.
Refine the answer with follow-up questions, filters, date ranges, packages, or company names.

Example questions:

"Which companies showed the strongest intent this week?"
"Summarize Docker pull activity for the last 30 days."
"Which companies are investigating but not yet in production?"
"Create a filter for Fortune 500 companies in investigation or experimentation."

Capabilities include:

natural-language answers over your Scarf analytics
company, package, web, and telemetry summaries
filter-aware analysis
follow-up questions that keep the current context
links back into Scarf where applicable

Each submitted AI question or follow-up that requires Scarf to process data may consume Runs.

Scarf Slack Agent

The Scarf Slack Agent brings Scarf AI into Slack. It is available on Basic and Premium plans. The Slack Agent is currently set up manually by the Scarf team in a shared Slack Connect channel with your team. If you do not already have a Slack Connect channel with Scarf, reach out and Scarf can help set one up.

To set it up:

Reach out to Scarf and ask for the Slack Agent to be enabled for your organization.
Scarf will confirm your plan, organization, and Slack Connect channel details.
Scarf will manually add and configure the Slack Agent in the shared channel.
Ask Scarf questions directly in Slack.

Example uses:

ask for a daily or weekly summary of account activity
investigate a company before a sales or success conversation
ask which packages or docs pages are driving interest
share Scarf context with teammates without switching tools

Capabilities include:

natural-language Scarf analytics in Slack
answers scoped to your Scarf organization
team-visible investigation threads
downloadable reports sent back into Slack
scheduled recurring tasks, such as cron-style reports or monitoring workflows
the same Scarf data access controls used by your organization

The Scarf Slack Agent consumes 1 Run per message sent to the AI. API calls the agent makes while answering that message do not consume additional Runs.

Scarf AI via API

You can also build Scarf-powered AI workflows with the Scarf API. This is useful when you want to route Scarf data into your own applications, internal agents, enrichment pipelines, or reporting systems.

To use the API:

Create or obtain a Scarf API token for your organization.
Store the token securely as an environment variable or secret in your application.
Call the relevant Scarf API endpoints for the data your AI workflow needs.
Pass the returned data into your application or model with the context and guardrails your workflow requires.

For API details, see the Scarf API documentation.

Common API-backed AI workflows include:

enriching account records with Scarf usage signals
summarizing package or company activity for internal tools
monitoring suspicious or unexpected download activity
generating lead, customer-success, or developer-relations summaries
combining Scarf signals with CRM or product data

Capabilities include:

programmatic access to Scarf analytics
organization-scoped authentication
access to supported Scarf data and workflow surfaces
integration with your own agents, data pipelines, and reporting systems

Direct public API calls may consume Runs as described in Billing and Pricing. API calls made by Scarf AI while answering an in-app or Slack Agent message do not consume additional Runs beyond the 1 Run for the message.

Using Scarf in your LLM of choice

Scarf Skill

The Scarf Skill is an agent skill that teaches compatible AI assistants how to answer Scarf analytics questions safely using the Scarf API. Use it when your assistant supports skills and you want a guided Scarf workflow inside that assistant.

To set it up:

Download or clone the Scarf Skill from github.com/scarf-sh/scarf-skill, or point your LLM client directly at that repository if it supports installing skills from GitHub.
Install or enable the skill in your LLM client using that client's skill setup flow.
Provide a Scarf API token to the assistant as SCARF_API_TOKEN using your assistant's secret-management mechanism.
Tell the assistant which Scarf organization or owner to use.
Ask Scarf analytics questions in natural language.

Example prompts:

"Using Scarf, summarize company activity for my organization over the last 30 days."
"List high-intent companies from this saved filter."
"Create an ad hoc insights filter for enterprise prospects in investigation."

Capabilities include:

guided use of supported Scarf API capabilities
safe defaults such as organization scoping and recent date ranges
read-oriented analytics flows
reusable analysis patterns where supported
concise summaries designed for operators and go-to-market teams

Skill-driven Scarf API calls consume Runs when they call Scarf endpoints or ask Scarf to process analytics.

Scarf MCP server

The Scarf MCP server lets MCP-compatible clients, such as desktop assistants and coding agents, connect to Scarf analytics as tools. It is available at:

https://api.scarf.sh/mcp

Prerequisites:

a Scarf API token
an MCP-compatible client that supports remote MCP servers

Configure your MCP client to use Scarf's remote MCP endpoint and authenticate with your Scarf API token. For clients that use JSON MCP configuration, add a server entry like this:

{
  "mcpServers": {
    "scarf": {
      "url": "https://api.scarf.sh/mcp",
      "headers": {
        "Authorization": "Bearer your-token-here"
      }
    }
  }
}

Restart your MCP client after saving the configuration, then ask it Scarf questions.

Example prompts:

"Use Scarf to summarize activity for this company."
"Find high-intent companies for my organization this month."
"Show me what changed in our Scarf activity this week."
"Help me investigate unusual download activity."

Available MCP capabilities include Scarf analytics, company and package insights, filtering, and related workflows exposed by the Scarf MCP server. The exact tool set may change over time, so rely on your MCP client to show the current available tools.

MCP tool calls use the Scarf API, so direct API-backed requests may consume Runs.

Choosing the right option

Use Scarf AI in-app when you are already working in Scarf.
Use the Scarf Slack Agent when your team wants shared analytics conversations in Slack.
Use the Scarf API when you are building a custom workflow or internal product.
Use the Scarf Skill when your assistant supports skills and you want a guided Scarf-aware workflow.
Use the Scarf MCP server when your LLM client supports MCP tools.

If you are not sure which setup is right for your team, contact Scarf and describe where you want Scarf answers to show up.