Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.honeycomb.io/llms.txt

Use this file to discover all available pages before exploring further.

​
Overview

Both Honeycomb and Embrace provide MCP servers that give AI assistants direct access to your observability data. When you connect both servers to the same agent, you can investigate the full request path—from mobile session data in Embrace to backend traces in Honeycomb—without leaving your development environment.

​
Before you begin

Before you get started, make sure you have:
Before running a combined investigation, confirm both servers are connected by asking your agent: “List the tools available from Honeycomb MCP and Embrace MCP.” If either server is missing, revisit Agent Skills or Embrace MCP setup before continuing.

​
How it works

With both MCP servers connected, your AI assistant has access to tools from each platform simultaneously. Honeycomb MCP gives your agent access to backend traces, BubbleUp analysis, and query execution. Embrace MCP gives it access to mobile session data, crash and ANR details, and network endpoint performance. The trace.id and emb.dashboard_session attributes on forwarded spans connect the two: your agent can follow a trace ID from Honeycomb into the originating Embrace session, or follow a session ID from Embrace into the corresponding backend trace in Honeycomb. Your agent handles the pivot between servers automatically within a single conversation. You don’t need to copy IDs or switch tools yourself.

​
Example investigations

The examples below are illustrative. Adapt the service names, app versions, and time windows to match your environment.

​
Triage a backend latency spike with user context

Ask your agent to investigate a latency spike in a specific service and determine whether real users are affected:
“There’s a latency spike in the checkout-api service in the last hour. Use Honeycomb to identify which endpoint and dimension is responsible, then use Embrace to tell me how many users are affected and what their experience looked like.”
The agent can use Honeycomb MCP to run BubbleUp on the slow spans, identify the top differentiating dimension, and retrieve a representative trace ID. It can then use Embrace MCP to look up the session associated with that trace and report on user impact, UX score, and conversion data.

​
Diagnose a mobile ANR with a backend root cause

Ask your agent to investigate an ANR spike and determine whether the cause is on the client or upstream:
“Embrace is showing an ANR spike on the checkout screen in the last 30 minutes. Find out what’s blocking the main thread and whether the backend is responsible.”
The agent can use Embrace MCP to retrieve the ANR breakdown, identify the in-flight network request, and extract the trace.id from the forwarded span. It can then use Honeycomb MCP to open that trace, walk the waterfall, and run BubbleUp to name the culprit backend service or dependency.

​
Check the health of a release across both platforms

Ask your agent to compare mobile and backend health before and after a deployment:
“We deployed version 7.15.0 of the mobile app this morning. Compare crash rates, ANR rates, and backend latency for checkout-api before and after the deploy.”
The agent can use Embrace MCP to retrieve crash and ANR metrics segmented by app version, and Honeycomb MCP to query backend latency for the same time window, surfacing any correlated degradation.

​
Writing effective prompts

The more context you give your agent, the less time it spends figuring out where to look. Include:
  • The service name, dataset, or environment in Honeycomb
  • The app name or app ID in Embrace
  • The time window you want to investigate
  • The specific metric or signal you are concerned about
To learn more about prompting agents effectively with Honeycomb MCP, visit Honeycomb MCP Use Cases.