- LATENCY
- COMPARISON
- USAGE
- HUMAN
Short answer
How can I check model performance with my AI?
You can ask your AI to pull specific metrics like latency or request counts for any model in your catalog. It will use the connected data to identify spikes and compare how different models are behaving. You'll get a clear breakdown of performance trends and specific model comparisons.
Where every insight lands.
Performance outcomes
Your AI turns raw request data into actionable performance reports.
LATENCY
Identify latency spikes
The AI scans request logs to find where response times jumped. You get a list of specific timestamps or models causing delays.
COMPARISON
Compare model metrics
The AI pulls stats for different models in your catalog. You can see which ones are faster or more reliable side-by-side.
USAGE
Track request volume
The AI counts how many requests each model handles. This helps you see which models are carrying the most load.
HUMAN
Manual tuning
The AI provides the data, but you decide which models to swap or upgrade. You handle the actual configuration changes in your environment.
The workflow
What your AI does when the data arrives.
The AI acts as a data analyst for your model telemetry.
Inventory models
The AI first checks which models you have available in your catalog to understand the scope of the analysis.
list_modelsFetch request logs
It pulls the specific history of requests for the models you want to investigate.
list_requests_by_modelCalculate metrics
The AI processes the request data to find patterns in latency or error rates.
get_analyticsSummarize findings
The AI presents the final comparison or spike report in plain language for you to review.
get_analytics
Try it
Copy these to start.
Use these prompts to begin your analysis.
Starting points
These are starting points. Swap out the model names or timeframes to match your specific needs.
Keywords AI Connector
You're all set. Choose your MCP client and follow the setup instructions.
https://edge.vinkius.com/vk_preview_R4VkEIrwLVvP7dLw0Fdy9vfjO7WqxWYeI2hDwlBU/mcpClaude Desktop
Follow the steps below to connect in seconds.
- 1In Claude Desktop, open Settings → Connectors.
- 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
- 3Click Add and start a new chat — Keywords AI capabilities are ready to use.
{
"mcpServers": {
"keywords-ai-mcp": {
"url": "https://edge.vinkius.com/vk_preview_R4VkEIrwLVvP7dLw0Fdy9vfjO7WqxWYeI2hDwlBU/mcp"
}
}
}Which models are currently in my catalog?
Show me the latency trends for the GPT-4o model over the last 24 hours.
Compare the average response times between my Llama and Claude models.
Did any models show a latency spike in the last hour?
Claude
ChatGPT
Cursor
VS Code
Windsurf
Claude Code
JetBrains
Cline
Start here
Connect Keywords AI once, then ask.
Just follow the one-click connection process. Your credentials stay encrypted, and you can immediately start asking questions about your model performance.
Connect Keywords AI to your AIFAQ
How this task behaves
- 01
Can the AI change my model settings?
No. The AI can only read analytics and request data. You must make any configuration or model changes yourself in your own dashboard.
- 02
Can the AI delete old request logs?
No. The tools only allow the AI to list and retrieve data, not to modify or delete it.
- 03
Does this work with any model?
It works with any model that appears in your Keywords AI catalog.
- 04
Can I see error rates?
Yes, the AI can pull analytics to show you how often requests are failing.
- 05
How often is the data updated?
The AI pulls the most recent data available through the Keywords AI API at the moment you ask the question.
More questions about Keywords AI? The Connector page answers them. See everything the Keywords AI Connector can do
Connect Keywords AI to Claude, Cursor, ChatGPT & more
