> ## Documentation Index
> Fetch the complete documentation index at: https://docs.originhq.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Ask Origin
> Natural language queries over your AI observability data — ask anything about your fleet.
**Ask Origin** is the AI assistant built into the Origin console. It has direct access to your fleet's AI observability data and can answer questions in natural language — no SQL, no dashboard configuration required. Every answer is grounded in your actual endpoint data.
## Where to find it
Ask Origin lives in the **right panel** of the Origin console. It's always accessible regardless of which view you're in (Overview, X-Ray, Inventory, Canvases). Click the **New chat** button at the top of the right panel to start a fresh conversation.
## Starting a conversation
### Using pre-built prompts
When you open a new chat, Ask Origin surfaces contextual prompt suggestions tailored to your current data. These are starting points, not limitations. Examples you'll see:
* **MCP tool sprawl over time** — *"Show me the top MCP tools over time by the number of distinct endpoints using each one, in weekly buckets over the last 8 weeks, so I can see which servers and tools are spreading across the fleet."*
* **Skills running in the field** — *"Show me the top agent skills running in the field over time, sampled daily over the last 30 days, so I can see which skills are active and spot any that appear or drop off across the window."*
* **Where token capital went** — *"Show me where token capital was spent in the last 7 days by provider, model, client surface, endpoint, and registered user."*
* **New since last week** — *"Compare this week to last week and show me everything new: MCP servers, tools, models, providers, and endpoints seen for the first time."*
* **How concentrated is your AI usage** — *"Show me how concentrated AI usage is across the tenant over the last 30 days — what share of all prompts and tool calls comes from the top endpoints and the real registered users behind them."*
Click any suggestion to run it immediately.
### Typing your own query
Click in the input box at the bottom of the right panel and type any question about your AI fleet. Ask Origin understands:
* **Metric queries** — *"How many sessions did Claude have last week?"*
* **Comparison queries** — *"Compare Claude usage this month vs last month"*
* **Anomaly queries** — *"Which endpoints have unusually high token spend?"*
* **Inventory queries** — *"What AI agents are running on endpoints that don't have Copilot?"*
* **Trend queries** — *"Is AI usage growing week over week?"*
## How Ask Origin generates answers
Ask Origin uses a two-step process:
1. It translates your natural language question into a **Kernel** — a structured analytic query against your Origin data
2. It executes the Kernel and returns a visualization (chart, table, or metric card) alongside a plain-language explanation
The generated Kernel is shown alongside each answer. You can fork it into your Library to save it for future use.
## Working with answers
After Ask Origin responds, you can:
* **Drill down** — click elements in the visualization to filter or narrow the query
* **Fork to canvas** — add the visualization to a custom Canvas
* **Save to library** — save the underlying Kernel for reuse
* **Follow up** — ask a follow-up question in the same thread; Ask Origin maintains context across the conversation
Ask Origin is context-aware. If you've scoped the console to a specific agent or endpoint group in the left panel, Ask Origin's answers will reflect that scope automatically — you don't need to repeat the filter in your question.
## Tailoring suggestions to your org
Click the **"Tailor these to your data"** link at the bottom of the suggestions list to profile your organization's AI usage patterns. This customizes the pre-built prompt suggestions to be more relevant to your specific fleet composition and usage patterns.