dog person. product @datadoghq, mcp & agent exp. prev @microsoft, @harvard 🇦🇺
New York, NY
Joined June 2017
- Tweets233
- Following312
- Followers191
- Likes1.7K
Amy Zhou retweeted
Your agents got some new tools this week, and so did you: Bits AI in Claude Code and Codex, voice agent observability, smarter RUM retention, LLM eval coverage, and a whole lot more.
Here's what shipped at @datadoghq this week:
- Bits AI can now be called from Claude Code, Codex, or any custom agent: ask questions, trigger and steer investigations, and read the results without leaving your editor.
- AgentObs Voice: listen to your AI voice agent's conversations directly in Datadog, with ready-made audio and image evaluators.
- Synthetic Monitoring & Testing Overview page is live: see what's unhealthy and what's not tested, on one page.
- The EC2 Spot Instance Assessment Tool (Preview) analyzes every Auto Scaling Group to find which workloads can actually run on Spot. (@mauerbac, @astuyve)
- Bits Code now works with Bitbucket Cloud Premium: get a tested PR back in your own repo, no migrating to GitHub required. (@diamondbishop)
- Suggested RUM retention filters calibrate session retention using your own application's performance data.
- End User Device Monitoring is now GA: find and investigate device issues before a ticket gets filed.
- Remap environments from the UI (APM, Preview): rename, merge, or split env values with no code changes.
- Patterns in LLM Observability finds the gaps between your eval datasets and real production traffic.
- The documentation toolset in the Datadog MCP Server is GA: your agent can now search and cite docs.datadoghq.com directly (@amymzhou)
Plus, we're hiring for 2 roles in software delivery and AI.
Full write-up, docs, and links to everything we shipped ↓
github.com/DataDog/shipped/d…
don't miss your chance to see ALQ's exquisite sense of style in person!!!
We're running INFER/CON next week in SF
Think putting AI into prod. Getting inference to work. Performance. Costs. Long-running agents... Hard tech rather than high-level commentary.
The line-up is stellar, so I expect this to be quite good :-)
1
6
327
now that I'm no longer on the dataviz team, I can properly glaze them while feeling unbiased...like how pretty is our new heatgrid widget 🤩🌅!!!
Claude connects automatically. Production investigations go multiplayer. Mobile profiling gets down to the CPU method.
And that’s just the beginning of what shipped this week at @datadoghq Here's the full rundown👇
- Connecting an AI client to a tool usually means everyone authorizing separately, every time. Datadog auto-connect for @claudeai (beta) fixes that: admins configure once, teammates connect automatically, no sign-in screens, no identity provider needed.
- Datadog Canvas (Preview) turns production investigations multiplayer: explore, investigate, and keep evidence connected on one shared view with teammates and AI agents.
- Datadog Agent Observability is now live on the Claude Marketplace, with Datadog as a featured launch partner.
- 3 platforms RUM couldn't fully see before now have official SDKs: C/C++, Electron, and .NET MAUI.
- Time to Initial Display tells you a launch is slow. It doesn't tell you why. RUM Mobile Profiling attaches CPU profiles from real iOS/Android sessions so you can see which methods consumed CPU time.
- A lineage graph is only as strong as its weakest link, usually a custom script nobody instrumented. OpenLineage support in Jobs Monitoring closes that gap for any data processing job.
-Comparing 10+ services in a timeseries chart gets crowded fast. The new Heatgrid widget makes that comparison legible in dashboards and notebooks.
- Your dashboards just got more room to breathe: iPhone Duo support for Datadog Mobile. Unfold for expanded dashboards and incident timelines, fold to pick up where you left off.
which reminds me, I overheard at HH last night from an engineer --> "please tell your dataviz team I loveee the colors, the blue/purple/yellow combo on dashboards helped me feel calm during the wfh days of covid"
1
3
171
we're finally talking about Datadog Canvas 😍 !! this is built for people who are tired of reading and want to see pretty graphs to help understand what is going on
Production investigations, now multiplayer...
Connect telemetry, evidence, and next steps on one shared canvas with teammates and AI agents.
Meet Datadog Canvas. Reva shows how it works 👇
Sign up for preview: canvas.datadoghq.com/?utm_so…
@datadoghq
Amy Zhou retweeted
Users can now auto-connect plugins when they log in to Claude, with MCP servers that accept Claude as a trusted issuer.
If you’re building plugins for Claude, add support so users get started faster and stay connected without re-authorizing.
Dan joining me on my influencer arc, showcasing the power of Scout 1.0 😎😎
A database is struggling. The root cause? A node rotation two hours earlier.
See how Scout 1.0 powers Datadog Bits Investigation to follow the evidence from symptoms to root cause.
Comparable performance at 3–8x lower inference cost on our benchmark - @datadoghq 👇
1
3
338
Amy Zhou retweeted
A database is struggling. The root cause? A node rotation two hours earlier.
See how Scout 1.0 powers Datadog Bits Investigation to follow the evidence from symptoms to root cause.
Comparable performance at 3–8x lower inference cost on our benchmark - @datadoghq 👇
watch this space!! —-> @TobinSouth the space being Tobin’s twitter because he’s about to pop off and become a mega influencer I just know it ;)
1
2
134
