> ## Documentation Index
> Fetch the complete documentation index at: https://symphony-docs.fcamara.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Overview and Runs
> Track your company's AI usage: active users, tokens, costs, models, agents, and every run.
**Restricted access.** These screens are intended for users with the **Administrator** (or **Master**) profile. Both are in **beta**.
## What it is
The **Overview** and **Runs** screens show **how your company is using AI**: how many people use it, how much they use it, which models and agents are most popular, how much it costs, and whether there were errors.
* **Overview** is the monitoring dashboard, with summary numbers and charts. Use it to answer questions like "is usage growing?", "which area uses it most?", or "how much did we spend this month?".
* **Runs** is the detailed list of **every AI use**: each question answered in chat and each flow step. Use it to investigate a specific case, like "why did this user's response fail?".
## How to access
User menu → **Admin Panel**.
**Money** values (cost) only appear if your company has the **view costs** permission in the Admin Panel. Without it, the screens show tokens and runs, but not cost.
## Basic concepts
| Term | What it means |
| - | - |
| **Run** | Each time AI was used: a chat response, a flow step, etc. |
| **Tokens** | How AI measures text size. Roughly, each word equals one or two tokens. More tokens means more usage and higher cost. |
| **Input tokens** | What was **sent** to the AI: the question, the conversation history, and attached files. |
| **Output tokens** | What the AI **generated** in the response. |
| **Latency** | How long the AI took to respond. |
| **Active users** | People who used AI at least once in the period. |
## Filters
At the top of both screens you choose:
* **Period** — **Last 24 hours**, **Last 7 days**, **Last 30 days**, **Last 90 days**, or **Custom** (pick the dates).
* **Business unit** — all BUs or a specific area.
All numbers on the screen change based on the filters.
## Overview
### Key indicators
| Indicator | What it shows |
| - | - |
| **Active users** | How many people used AI in the period. |
| **Total tokens** | Total volume of text processed. |
| **Total cost** | How much was spent in the period *(only with the view costs permission)*. |
| **Total runs** | How many times AI was used. |
Each indicator shows the **change** compared to the previous period of the same length. For example, for the last 7 days, the comparison is with the 7 days before.
### Usage over time
Day-by-day usage chart. In **Metric**, switch between **Tokens**, **Runs**, and **Cost**.
### AI consumption
The most used **Models** and **Agents** in the period, with the tokens and (with permission) the cost of each. It helps you see which agents are delivering value and which are barely used.
### Organizational consumption
The **Users** and **Business units** that used AI the most in the period.
### Recent runs
The latest runs, with the same filters as the screen. To see all of them, use the **Runs** screen.
## Runs
The **Runs** screen lists every AI use, with these columns:
| Column | What it shows |
| - | - |
| **Status** | **Success**, **Warning**, or **Error** (see below). |
| **When** | Date and time of the run. |
| **User** | Who used it. |
| **Business unit** | The person's area. |
| **Agent** | The agent used, if any. |
| **Model** | The AI model that responded. |
| **Latency** | How long the response took. |
| **Input tokens** and **Output tokens** | The size of what was sent and what was generated. |
| **Total cost** | The cost of the run *(only with the view costs permission)*. |
### What each status means
| Status | Meaning |
| - | - |
| **Success** | The AI responded normally. |
| **Warning** | The AI responded, but something was off. For example, the chosen model failed and Symphony **automatically switched** to another (see [Automatic model fallback](/en/admin/painel-administrativo#automatic-model-fallback)), or the response arrived without a token count. |
| **Error** | The AI couldn't respond. |
### Search and filter
* Use **search** to find runs by user name or email, model, or agent.
* Click the **Status**, **User**, **Business unit**, **Agent**, and **Model** column headers to filter by one or more values. Active filters appear in a bar above the table. Remove them one by one, or all at once with **Clear filters**.
* Click a column header to sort.
### View run details
Click a row to open **Run details**, with latency, tokens, cost, user, agent, model, date and time, and a **preview of the text sent** to the AI.
**Mind privacy.** The preview shows part of what the user wrote. Treat this information with the same care as any personal or confidential company data.
### Export
Click **Export** and choose **Export CSV** or **Export Excel**. The export uses the same filters and search that are active on the screen.
### Delete a run
Use the **Delete** icon on the row. The run no longer appears in the list or in the totals.
## Usage examples
| Question | Where to look |
| - | - |
| How much did we spend on AI this month? | **Overview** → **Last 30 days** period → **Total cost**. |
| Which area uses it most? | **Overview** → **Organizational consumption** → **Business units**. |
| Is the agent we created being used? | **Overview** → **AI consumption** → **Agents**. |
| A user reported an error yesterday. | **Runs** → search by email → filter **Status** = **Error** → open the details. |
| I need a report for leadership. | **Runs** → adjust the filters → **Export Excel**. |
## Best practices
* **Check the Overview every week** to spot changes in usage or cost early.
* **Compare equal periods.** The indicator change already does this for you.
* **Investigate Error spikes** on the Runs screen: filter by **Status = Error** and check whether the problem is concentrated in one model or agent.
* **Use export** instead of copying numbers by hand.
* **Set model prices** in **Settings → Models**. Without a price, run cost shows as zero.
## Frequently asked questions
Cost is only calculated for models with a **price configured** in **Settings → Models**. Models without a price show zero cost, not an estimated value.
Your company doesn't have the **view costs** permission in the Admin Panel. Contact the platform manager.
Yes. It happens when the attempt failed before the AI started responding (**Error** status), or when the response arrived without a token count (**Warning** status).
Yes. A temporary chat isn't kept in the user's history, but its AI usage is counted.
Yes. Search by name or email on the **Runs** screen, or open the user's details in **Management → Users**.
## Known limitations
* The screens are in **beta** and may change.
* Cost depends on the price configured for each model.
* Cost values require the **view costs** permission.
* The preview of the sent text is a summary and doesn't show the whole conversation.
* Peak usage time and dates appear in your computer's time zone.
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.