GIVE YOUR AI AGENT EYES ON PRODUCTION.

PixelGenome connects live OpenTelemetry traces directly to Cursor, Windsurf, Claude, and Codex via MCP—so your agent can find, diagnose, and fix production bugs without pasted stack traces.

$ npm install @pixelgenome/node-sdk

NODE.JSSUPPORTEDPYTHONCOMING SOON

WORKS WITHCURSORWINDSURFCLAUDECODEXANY MCP CLIENT

LIVE AGENT SESSION

CURSOR / AGENTsrc/services/stripe.ts

> Fix the 500 on /api/pay from 5 minutes ago.

01
get_latest_errors()
FOUND #tr-9942
02
get_stack_trace("tr-9942")
stripe.ts:42
03
Editing local source
FIX READY

- pool.max = 32+ pool.max = 64

01 / PROTOCOL

OBSERVABILITY YOUR AGENT CAN CALL.

Not another dashboard. A typed tool surface built for autonomous debugging.

01get_latest_errors()

Find uncaught exceptions and failing routes.

↗

02get_trace_waterfall(trace_id)

Inspect the entire request path, span by span.

↗

03get_stack_trace(error_id)

Return only the frames that matter.

↗

04suggest_code_location(trace_id)

Map production failures to local files and lines.

↗

02 / AGENT LOOP

FROM INCIDENT TO PATCH, WITHOUT THE COPY-PASTE.

OBSERVE

Ingest standard OTLP traces, logs, spans, and exceptions from production.

INVESTIGATE

Your agent calls PixelGenome tools from the IDE and follows the trace to the root cause.

PATCH

Source-mapped evidence points the agent to the exact local file and line.

VERIFY

Trace the fix and the agent’s own tool calls with OpenTelemetry.

03 / AGENT RESPONSE

ONLY THE CONTEXT YOUR AGENT NEEDS.

When your coding agent calls PixelGenome, it receives a compact, structured response—not thousands of raw log lines. Each response includes the failure, trace ID, relevant stack frames, and source location.

STRUCTURED JSONRELEVANT FRAMESSOURCE MAPPED

{
  "error": "PoolTimeoutError",
  "route": "POST /api/pay",
  "trace_id": "tr-9942",
  "source": {
    "file": "src/services/stripe.ts",
    "line": 42
  },
  "evidence": {
    "db_pool_active": 32,
    "db_pool_max": 32
  }
}

04 / LOCAL SETUP

YOUR STACK. YOUR AGENT. ZERO GUESSWORK.

Connect your telemetry, point the MCP server to live traces, and give your agent production-grade context before it ships a patch.

AI_SETUP.sh● ● ●

$ docker compose up -d

✓ telemetry ready on :4318

✓ trace store ready on :8123