Working with Datadog and AI In-Platform and Outside

There are many ways to put AI to work with Datadog:

Use Bits in-platform, bring your own harness through the Datadog MCP server, or work directly against the API.

All three can be useful. They also come with very different levels of setup, reach, maintenance, and control. Treating them like interchangeable ways to “add AI” is a good way to end up with more plumbing than problems solved.

Datadog and NoBS engineers will explain where each approach fits, what you trade for that flexibility, and where a human still needs to make the call.

We’ll run through a real workflow: investigate a problem or make a change, query Datadog to see what actually happened, correct what went wrong, and turn the parts worth repeating into a skill or standing automation.

What to Expect:

• When to use Bits, MCP, or direct API access, and what you give up with each
• How much context and access an agent needs to do useful work
• How to use agents for investigation, instrumentation, and Datadog resource creation, then verify the result against real telemetry
• When you’ve typed the same prompt enough times that it should probably become a skill or standing automation

Why Should You Attend?

Because production does not care that the prompt looked convincing. You should know what happened, whether it worked, and what still needs a human.