AI agents · MCP
Early accessSession data, ready for your AI agents
uxTrace prepares problem sessions in a format AI agents can read. Agents review them, find the causes of the problems users face and answer with links to the sessions. Use it with cloud models such as Claude or ChatGPT, or local models such as Qwen.
Capabilities
What your agents can do
- !Find problem sessionsRage clicks, console errors, failed requests and other anomalies, found for you.
- ▶Review sessions in an AI-ready formatActions, events, console and network on one structured timeline, prepared for the model.
- ?Determine the causesAgents explain what went wrong and why users get stuck, with the sessions as proof.
- >_Summarize errors and requestsTop errors, failing endpoints and affected users.
- Search sessions and pagesBy user, URL, device or event; sessions and clicks for any page.
Why couldn't users save the form yesterday?
uxTrace · search sessionserrors on /forms, last 24 hours → 37 sessions
uxTrace · read session timeline3rd click on Save → POST /api/forms/save · 500
37 sessions failed with a 500 on
POST /api/forms/save after the third click on Save. The console shows TypeError reading 'id'. Most affected: Chrome on Android.
Session #47Session #155Session #157
How it works
Ask in plain language, get answers with sessions
- Connect. Add the uxTrace MCP server to your AI agent or assistant.
- Ask. Describe the problem or question in plain language.
- Get the cause. The agent reviews the problem sessions and replies with the cause and links to the sessions.
Models
Cloud or local models
Choose where your AI runs. With on-premise uxTrace and a local model, the whole setup stays inside your network.
Cloud modelsClaude, ChatGPT and other MCP-compatible assistants connect to uxTrace in our cloud or on your server.
Local modelsQwen and other open models run next to on-premise uxTrace. No data leaves your network.
| uxTrace | AI model | Where session data goes |
|---|---|---|
| Cloud | Cloud model | uxTrace Cloud and your model provider |
| On-premise | Cloud model | Your servers and your model provider |
| On-premise | Local model | Stays inside your network |
Example prompts
Built for support, engineering and product teams
Support
Show the last sessions of user 7f3c9a1e and what went wrong.The agent finds the sessions by User ID and points to the error.
Engineering
Which endpoints returned 5xx most often this week, and on which pages?A summary of failing requests with sessions to replay.
Product
Find sessions with rage clicks on the checkout page yesterday.Abnormal sessions with page stats and heatmap data.
FAQ
Questions about AI agents and MCP
- What is the uxTrace MCP server?
- A Model Context Protocol (MCP) server that gives AI agents structured access to uxTrace session data: sessions, timelines, console errors, network requests, page stats and heatmaps.
- What can an agent find out?
- Agents review problem sessions in an AI-ready format (actions, events, console and network on one timeline), determine the causes of the problems users face and answer with links to the sessions as proof.
- Which AI models can use it?
- Cloud models such as Claude or ChatGPT and other MCP-compatible assistants, and local models such as Qwen and other open models.
- Can the whole setup run without data leaving our network?
- Yes. With on-premise uxTrace and a local model, session data stays inside your network.
- Do agents see masked data?
- No. Agents see only what uxTrace stores; fields masked on the user's device stay masked.
- Is it available now?
- The MCP server is in early access for pilot clients.
Early access
Connect your AI agents to uxTrace
The MCP server is available to pilot clients. We'll set it up with your cloud or local model.