> For the complete documentation index, see [llms.txt](https://docs.cortex.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.cortex.io/get-started/cortex-ai-assistant.md).

# Chatting with Cortex

Cortex answers natural-language questions about your engineering data—services, ownership, Scorecards, Initiatives, incidents, deployments, and more—so you can get answers without digging through the UI yourself. You can chat with Cortex wherever you work:

* **In Slack via the AI Assistant** - Mention `@Cortex` in a channel or DM to ask questions in plain language. See [Interacting with Cortex in Slack](/ingesting-data-into-cortex/integrations/slack/using-the-integration-for-slack-ai-assistant.md#interacting-with-cortex-in-slack)
* **In your AI tools via MCP** - Connect the [Cortex MCP](/get-started/cortex-ai-assistant/mcp.md) to ask questions from the tools your team already uses.

However you ask, answers are grounded in your real Cortex data and respect each individual's permissions. The rest of this page covers what you can ask and how to write effective prompts; see the surface-specific guides for setup.

Here are a few prompt examples for you to get started with:

* ***What are the patterns of incidents on my team over the last 2 quarters? What areas do you recommend we invest in to improve the reliability of our services?***
* ***Give me a summary of everything I worked on and my key accomplishments over the last quarter.***
* ***Has there been an increase in PR sizes recently? Does that correlate with longer review times?***
* ***What is our deployment frequency for the last 30 days across the services I own? Are there any patterns that may be causing that to change compared to the last quarter?***

See the [Cortex AI prompt library](/get-started/cortex-ai-assistant/library.md) for a full list of prompt patterns organized by use case and role.

## Available capabilities

The following API endpoints are accessed when chatting with Cortex:

<table data-full-width="false"><thead><tr><th width="283.66015625">Capability</th><th>Description</th></tr></thead><tbody><tr><td><strong>Entity details</strong></td><td>Look up ownership, metadata, and custom data for any entity</td></tr><tr><td><strong>Entity descriptors</strong></td><td>View and list entity descriptors (YAML definitions)</td></tr><tr><td><strong>Dependencies</strong></td><td>List incoming and outgoing dependencies for an entity</td></tr><tr><td><strong>On-call</strong></td><td>Check who is currently on-call for an entity</td></tr><tr><td><strong>Scorecards</strong></td><td>List Scorecards, view scores, and get next-step recommendations</td></tr><tr><td><strong>Initiatives</strong></td><td>List and inspect Initiatives and their progress</td></tr><tr><td><strong>Relationships</strong></td><td>Explore entity relationships and relationship types</td></tr><tr><td><strong>Team details</strong></td><td>View team membership, sourcing, and contact channels</td></tr><tr><td><strong>Workspace context</strong></td><td>Query "my" entities and how they perform against Scorecards</td></tr><tr><td><strong>Eng Intelligence metrics / traces</strong></td><td>Query DORA metrics, incident trends, PR data, and other engineering performance indicators</td></tr><tr><td><strong>Deployments</strong></td><td>Search deployments across the workspace by environment, type, entity, and time range; list available deployment environments</td></tr></tbody></table>

## Providing feedback

Reaction emoji options are displayed after each response. You can react with `:thumbs_up:` (👍) or `:thumbs_down:` (👎) to provide feedback on the quality of the answer. This feedback helps improve the responses over time.

## Crafting effective prompts

Getting useful output from any AI assistant comes down to how you ask. Be sure to write prompts that are clear, specific, and designed to get you the response you need.

* **Be specific** - Instead of ***Tell me about services***, you could say, ***Show me all critical services failing the Production Readiness Scorecard***. This provides you with more actionable information.
* **Combine multiple questions when context helps** - Cortex handles complex prompts like ***Who owns orders-api, when was it last deployed, and are there any open incidents?*** You receive richer context in response to one question instead of piecing together three separate queries.
* **Reference your organization's specific constructs** - If you've defined custom Scorecards or Initiatives in Cortex, reference them by name. Cortex understands your organization's specific standards and can query against them directly.
* **Use follow-up questions to drill down** - Start broad, then go narrow. ***Show me services with failing Scorecards***, followed by ***What specific checks is checkout-service failing?*** moves you from landscape view to action items.
* **Build a library of your most-used prompts** - If you ask the same question every Monday morning, save it. Some teams maintain shared prompt libraries so everyone can benefit from patterns that work.

See the [Cortex AI prompt library](/get-started/cortex-ai-assistant/library.md) for a full list of prompt patterns organized by use case and role.