Your AI coding tool knows the code.
Now give it the user.

Uzera's MCP server pipes structured session replay, error context, churn impact, and fix history directly into your AI coding environment. One command to install. Zero context switching to debug.

START YOUR FREE TRIAL

AI coding tools are powerful.
They're just missing half the picture.

Integrations connect Uzera to the third-party platforms your team relies on, so every user action Uzera captures a error, survey response, a tour completion, a feature interaction—flows directly into your analytics dashboards, CRM records, spreadsheets, and communication channels.

What Your AI Tool Has
Feature name
Add in permission
Full codebase access
What the user actually did
Machine-readable session replay
Stack traces
Which users are at churn risk
Churn impact + ARR exposure
Git history
How your team fixed this before
Pattern memory + fix history
Test coverage
Whether this error spiked
Cross-signal correlation
Linting, types
When this version deployed
Automatic deploy detection

At-a-Glance Performance: Aggregate Dashboard Metrics

The Products page opens with four cross-product KPI stat cards for an immediate portfolio health check:

Total Products

all websites and apps connected to your account, displayed with a supporting tooltip

Total Sessions

aggregate session count across every product in your account

Active Users

total active users recorded across all connected products

Tracked Events

combined count of custom events being tracked across your entire account

8 purpose-built tools.

10× richer context per call.

Most MCP integrations send a surface-level summary. Uzera sends structured, machine-readable data built specifically for debugging.

#1

get_error_context

Returns the full error object: stack trace, affected endpoints, browser/device breakdown, and first/last seen timestamps.

     {        

          "error_id": "err_8x92kq", 

          " message": "TypeError: Cannot read proper of undefined",  
            "file": "cart.js", 
  
           "line": 847, "
    
           “affected_sessions": 23,  
            "affected_users": 19

     }

#2

get_session_replay

Returns the structured user journey — not a video URL. Machine-readable event sequence your AI tool can parse and reason about.

     {        

              "session_id": "sess_4hj29x",
             "events": [ { "type": "pageview", "url": "/cart" },,  
              { "type": "rage_click", "count": 4 },
             { "type": "error", "error_id": "err_8x92kq" }
          ]

     }

#3

get_churn_impact

Returns churn risk scores and ARR exposure for users affected by an error. Turns "how many users?" into "how much revenue?"

{          
         "affected_users": 19,    
         "churn_risk_breakdown":
     {
          high": { "count": 3, "arr_at_risk": 4200 }    
     }        
         "total_arr_at_risk": 12100
  
 }

#4

get_pattern_memory

Searches your team's resolved errors for the same pattern. Returns how you fixed it before and how long it took.

{ 
    
        "similar_resolved": 

         [
      
             { 

                  "resolved_at": "2025-11-14", 
        
                  "time_to_fix_hours": 3.5, 
    
                  "fix_summary": "Added null guard on cart.items" 
            }
       ]

}

#5

get_cross_signal_correlation

Links the error spike to concurrent signals: API latency changes, session drop-off, deploy events.

{ 
  
           "correlated_signals":

            [ 
      
                  { "type": "deploy", "version": "v2.14.1" },                     
  
                  { "type": "session_drop", "drop_pct": 34 } 
    
           ]

}

#6

get_deploy_detection

Returns the version history around the error window — automatically detected from your sessions.

{ 
  
               "current_version": "v2.14.1", 
  
               "error_rate_change": "+340%", 
                                   
              "regression": true

}

#7

get_funnel_drop

Returns conversion funnel data for the page where the error occurred — understand business impact beyond user count.

{ 
  
   "funnel": "checkout", 

     "baseline_completion_pct": 41.2, 
      "current_completion_pct": 16.5, 
  
   "delta": "-24.7pp" 

}

#8

get_web_vitals

Returns Core Web Vitals for the affected page. Flags performance regressions that may have contributed.

{ 

            "page": "/checkout", 
                                                          
            "lcp_ms": 4800, 

            "status": { "lcp": "poor" },

            "baseline_lcp_ms": 1900

}

PostHog's MCP: 25 generic tools.

Uzera's MCP: 8 purpose-built tools, 10× richer per call.

Signal
Post MCP
Uzera MCP
Session Reply
Link to open in browser
Structured event sequence—machine readable
Error Data
Basic Count + Message
Full trace, version correlation, deploy detection
User Impact
User Account
Churn risk score + ARR at risk per user
Fix History
None
Pattern memory from resolved errors
Linting, types
25(Generic Analytics)
8 (built for debugging)

Real-World Use Cases

Engineering — Fix in the IDE, not in 6 tabs

BEFORE MCP

  • Sentry alert fires
  • Open Sentry — read stack trace
  • Open FullStory — find session replay
  • Open Slack — search for thread
  • Open Linear — check if filed
  • Open Datadog — check API latency
  • Copy all to Claude Code manually
  • Ask for a fix

AFTER UZERA MCP

  • Sentry alert fires
  • Ask Claude Code: "What's happening on /checkout?"
  • Fix

Result: 40%+ reduction in mean time to resolution. Clear-trace bugs resolved in under 4 hours.

Product — Prioritize by revenue, not by volume

# Example prompt in Claude Code:

"Which open errors have the highest ARR at risk this week?"

Uzera MCP returns ranked errors with churn exposure.
Product and engineering 
can agree on what to fix next, with data behind it

Customer Success — Respond before the ticket arrives

Uzera MCP can be queried as part of a Slack alert workflow. When an error fires, the alert includes structured context your CS team can use to reach out proactively — before the user files a ticket.

Example Slack alert format:

⚠️ Error spike — /checkout — v2.14.1
Affected: 23 sessions · 3 users at high churn risk

ARR exposure: $4,200

Replay: [structured link]

Pattern: last seen Nov 2025 · fixed in 3.5h

→ Open in Claude Code

  • MCP Pricing

100 calls free. Scale when you need to.

Start free. Scale as you grow. No hidden fees.

Free Plan
Free for Lifetime
100
MCP Calls / Month
FREE
Choose Plan
Save 50%
Startup plan
Ideal for startups and small businesses
5000
MCP Calls / Month
$299 / Month
99
Choose Plan
Save 25%
Growth plan
Ideal for growing businesses
100,000
MCP Calls / Month
$299
179
/ Month
Choose Plan
Enterprise
Ideal for growing businesses
Unlimited
Ideal for growing businesses
Custom
Choose Plan
Starter
Free for lifetime
$0
/ Month
This Plan includes:
25K Sesion Replay
Product Analytics
Traffic Analytics
Heatmap Analytics
Form Analytics
Funnel Analytics
AI Agent Detection
User Journey
Product Tour
Checklist
Survey
DemoX
SegmentationS
Segments
1 Team Seats
1 Product
10 MCP
5gb Storage
Choose Plan
Save 15%
Growth
Free for lifetime
$106
125
/ Month
This Plan includes:
50K Sesion Replay
Product Analytics
Traffic Analytics
Heatmap Analytics
Form Analytics
Funnel Analytics
AI Agent Detection
User Journey
Product Tour
Checklist
Survey
DemoX
SegmentationS
Segments
1 Team Seats
1 Product
50k MCP
75gb Storage
Choose Plan
Save 15%
Pro
Free for lifetime
$237
299
/ Month
This Plan includes:
25K Sesion Replay
Product Analytics
Traffic Analytics
Heatmap Analytics
Form Analytics
Funnel Analytics
AI Agent Detection
User Journey
Product Tour
Checklist
Survey
DemoX
SegmentationS
Segments
1 Team Seats
1 Product
100k MCP
350gb Storage
Choose Plan
Enterprise plan
Ideal for multiple products or an enterprise
Let’s talk
This Plan includes:
Pro Plan incuded
Custom Plan
SLA
Dedicated Success Manager
Choose Plan

What counts as one MCP call:

Each tool invocation counts as one call. A single Claude Code query may invoke multiple tools — for example, get_error_context + get_churn_impact + get_pattern_memory = 3 calls.

Note: MCP calls are included in all Uzera plans. No separate MCP-only billing.

Built on the MCP open standard.
No lock-in. No vendor extension required.

Technical Specs

Protocol:

Model Context Protocol (MCP) — open standard by Anthropic

Auth:

Browser OAuth 2.0 — no API keys to manage

Transport:

SSE (Server-Sent Events)

Latency:

p95 < 400ms per tool call

Data freshness:

Errors and sessions: real-time · Churn scores: updated every 6 hours

Retention:

Governed by your Uzera plan (60 days on Starter → configurable on Enterprise)

Compatible clients:

Claude Code, Cursor, Windsurf, and any MCP-spec client

Security

All data is scoped to your Uzera workspace

OAuth tokens are short-lived and rotated automatically

No session replay data is stored by the MCP server — it proxies from Uzera's core platformGDPR read

SOC 2 compliant (Enterprise plan)

GDPR ready

Frequently Asked Questions

Does Uzera MCP work with Cursor and Windsurf, not just Claude Code?

Yes. Uzera MCP follows the open MCP specification. It works with any client that supports MCP — currently Claude Code, Cursor, and Windsurf.

What is an MCP call?

Each tool invocation is one call. When Claude Code queries Uzera for error context and churn impact together, that's two calls.

Is session replay data sent to my AI tool?

The structured event sequence is sent — not a video. Uzera converts the session into a machine-readable event log that your AI tool can parse and reason about directly.

Do I need to be on a paid plan to use MCP?

No. The Starter (free) plan includes 100 MCP calls per month. Paid plans include significantly more.

What happens to my data once it passes through MCP?

The MCP server proxies data from your Uzera workspace. It does not store data independently. Your Uzera data retention settings govern everything.

Can my whole engineering team use it?

Yes. MCP access is workspace-level. Any team member authenticated to your Uzera workspace can connect the MCP server to their own IDE.

Does this require changes to our codebase?

No. You need the Uzera SDK installed (one JS snippet, 5 minutes). The MCP server is separate and connects to your existing Uzera data.

Your next production bug already has a replay.
Give your AI coding toolaccess to it.

Uzera's MCP server connects session replay, error intelligence, and churn data directly to Claude Code and Cursor. One command. Zero context switching.