Repository navigation
Datadog Shipped - 11/13/26 #7
blondecostanza
announced in
Announcements
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Kicking off this roundup. 10 ships worth knowing about:
What shipped, why it exists, docs, and more.
1. Investigate issues with Bits from Claude Code, Codex, or any custom agent
What shipped: Bits AI can now be called from agents outside Datadog. Developers and SREs can ask questions, trigger and steer Bits Investigations, and read the results from the agent they already have open, through the Datadog MCP, API, or pup CLI.
Why it exists: Previously, working with Bits meant leaving whatever tool you were already using to open the Datadog UI. Now you can go from alert to fix faster: Bits starts investigating when a monitor fires, and your own agent can pull in the findings so you move straight to implementing the change. Conversations, findings, and evidence stay in Datadog, so teammates and agents build on the same analysis instead of repeating work in separate sessions.
How it works: Connect your coding agent using the Datadog MCP setup guide. Users need the Bits Chat Access permission and the Bits Investigation permission. Try a prompt like "Identify an alerting monitor, and investigate it with Bits to help me figure out the root cause and impact." Any Bits AI interaction over MCP, API, or CLI is priced through AI Credits, the same as in the web UI.
Shipped by: Daniel Green
Start now: Datadog MCP setup guide
Docs: Bits AI MCP docs · Bits AI API reference · Pup CLI · Release note
bits-ai-in-your-agents-with-elizabeth.mp4
Here's the prompt to try:
Identify an alerting monitor, and investigate it with Bits to help me figure out the root cause and impact
2. AgentObs Voice: listen to your AI voice agent's conversations directly in Datadog
What shipped: Agent Observability now supports audio and image evaluators. You can replay voice conversations and score audio and image quality with ready-made evaluators, directly in Datadog.
Why it exists: Multimodal agents, especially voice agents, have been a blind spot for most observability tooling, which focuses on text. Teams building voice agents need the same evaluation rigor they'd apply to text: catching quality issues and measuring whether a change actually helped, without building custom tooling to replay and score audio themselves.
Shipped by: Fouad Wahabi, Zach Groves
Docs: Audio and image evaluations · Blog post
riverside_magic_episode.01_elizabeth.s_studio.2.mp4
3. Introducing the Synthetic Monitoring & Testing Overview page: see what's unhealthy and what's not tested, on one page
What shipped: The Synthetic Monitoring & Testing Overview page is now generally available. From a single page, you can assess your monitoring status and coverage: see which tests are failing, spot patterns in failures, and discover high-traffic endpoints and views that aren't covered by a Synthetic test.
Why it exists: Customers need to know two things: what's broken, and whether they're testing what matters. Previously, answering either question meant building a dashboard or searching the test list by hand. The Overview page answers both without either.
How it works: A test health map shows where failures concentrate, groupable by location, type, priority, failure reason, environment, or team, with direct actions on each failing test (Investigate With Bits, View Traces, Run test now). A coverage gaps section shows untested API endpoints (from APM) and untested views (from RUM), ranked by real traffic over the past week, with high-traffic gaps highlighted and a one-click path to create a test directly from a gap.
Shipped by: Hiba Ijjaali
Start now: Synthetic Monitoring & Testing Overview
Docs: Overview page documentation
4. The EC2 Spot Instance Assessment Tool
What shipped: A tool, now in Preview, that analyzes every Auto Scaling Group across your AWS accounts and regions to identify which workloads are actually good candidates for EC2 Spot.
Why it exists: Spot can save up to 90% on EC2, but figuring out which workloads can actually tolerate running on it has always been the hard part. This tool does that analysis for you, based on current on-demand spend, estimated monthly Spot savings, workload behavior, and interruption risk.
Shipped by: Matt Auerbach
Start now: Request Preview access
Docs: Release note
riverside_edit_-.spot.instance.launch.video_elizabeth.s_studio.mp4
5. Use Bits Code with Bitbucket Cloud Premium
What shipped: Bits Code now supports Bitbucket Cloud Premium, generally available. Customers who host their code on Bitbucket Cloud can hand Bits Code a task or automation and get back a tested pull request in their own Bitbucket repository, no migrating or mirroring to GitHub required.
Why it exists: Bitbucket is the source control system for many large enterprises, and until now it blocked Bits Code there entirely. This also opens the Code Security fix motion on Bitbucket: SAST and IAST findings can now land as Bitbucket PRs, and any "Fix with Bits" step, including fixes Bits Investigation recommends, works there too.
How it works: Confirm your Bitbucket workspace is on Bitbucket Cloud Premium, install the Bitbucket Cloud Source Code integration and connect the workspace, then start a session at Bits AI > Code, or click "Fix with Bits" in a supported product. To run work on a schedule or in response to signals, create an automation pointed at a Bitbucket repository.
Shipped by: Diamond Bishop
Start now: Bits AI Code
Docs: Bitbucket Cloud Source Code integration · Bits Code supported providers · Release note
6. Calibrate RUM session retention with suggested retention filters
What shipped: For RUM without Limits customers, Datadog now shows which retention filters each RUM application should have, with recommended thresholds taken from that application's own performance data, now generally available.
Why it exists: Without this, customers didn't always know which sessions were important to keep or what a poor-performance threshold looked like for their own app, so they'd go through a slow, iterative process to create and refine filters, often retaining too many sessions in the meantime.
How it works: Curated suggestions appear per application across six categories (dynamic session sampling, errors, slow performance, user activity, sessions with add-on data, specific app versions). Performance-based thresholds use the 90th percentile for that view over a rolling 30-day window, refreshed every Monday. Selecting a suggestion opens a pre-filled filter form you can edit before saving; existing filters that match a suggestion are automatically checked off.
Shipped by: Arti Arutiunov
Start now: RUM retention filters
Docs: Suggested retention filters documentation · Retention filter best practices guide · Release note
7. End User Device Monitoring is now generally available
What shipped: End User Device Monitoring (EUDM) is now generally available. By extending the Datadog Agent to employee devices, EUDM brings device health, network, and application data into the same platform infrastructure and network teams already use, so IT can find and investigate device issues without waiting for a ticket.
Why it exists: Today, IT often learns about a device problem only after someone files a ticket, by which point a laptop may have been low on memory for weeks or a Wi-Fi connection may have been dropping calls for months. EUDM lets teams investigate without disrupting anyone and spot fleet-wide trends when one issue is affecting many people at once.
How it works: The Command Center surfaces auto-detected fleet-wide issues, prioritized by how many people they affect, which can sync to Jira or ServiceNow. Teams can see fleet health and OS breakdowns, review a timeline of notable events and configuration changes on a device, check battery and resource metrics, and connect choppy Zoom calls or a slow app back to the device, network, or VPN behind them through integrations with Network Path and Network Device Monitoring.
Shipped by: Erica Ho
Start now: End User Device Monitoring
Docs: Public documentation · Blog post · Release note
8. Remap environments from the UI
What shipped: Environment Remapping, now in Preview, lets you rename, merge, or split
envvalues directly in the UI, without any code or configuration changes.Why it exists:
envis one of the three reserved Unified Service Tags, and the primary way teams slice APM data by environment. When the same environment shows up under many different names (prod,Prod,prd,production), data gets fragmented: dashboards and monitors miss data, cost attribution gets complicated, and AI agents like Bits AI can miss relevant telemetry during root cause analysis. Cleaning this up at the source previously meant re-instrumenting and redeploying across every team involved; remapping rules fix it from the UI in minutes instead.How it works: Create a custom name to rename an environment 1:1 or merge several values under one standard name, or use a tag to split an environment into distinct subsets. Before creating a rule, you can see which monitors and dashboards reference the existing name, and preview the rule's effect on service-to-env cardinality.
Shipped by: Mariana Frangos
Start now: Request Preview access
Docs: Environment remapping documentation
9. Patterns in Agent Observability
What shipped: Patterns, the clustering feature in Agent Observability, automatically surfaces topics from your production span traffic, and can compare your offline eval datasets against that traffic to find coverage gaps, proactively surfacing spans that aren't represented in your evals.
Why it exists: Knowing what your agent's real production traffic actually looks like, and whether your eval suite actually covers it, has typically meant manual review. Patterns automates finding what's missing from an eval suite; a customer still fills in expected outputs before using a surfaced case to evaluate their agent.
How it works: In Patterns configuration, add your evaluation datasets, and Patterns automatically curates a new dataset holding the coverage gaps for each uncovered topic. Coverage analysis and suggested datapoints are visible per topic.
Shipped by: Vincent Cornet, with Greg Svigruha
Docs: Set up a Pattern · Find evaluation coverage gaps
riverside_edit_-.2026.11.05.18.57.40.1._elizabeth.s_studio.mp4
10. The documentation toolset in the Datadog MCP Server is now GA
What shipped: Your AI agent can now search and read docs.datadoghq.com directly through the Datadog MCP Server, so its answers come from current documentation, with links it can cite.
Why it exists: Without this, an agent's answers about Datadog came from its training data, which goes stale and can't cite a source. This toolset grounds answers in the documentation as it exists today.
How it works: If you use
?toolsets=all, you already have this. Otherwise, adddocumentationto your toolsets.Docs: Documentation toolset
Open roles
We have lots of open roles in the software delivery and AI space, here are a couple:
Staff Product Manager, Self Improving Software (New York)
Founding product work on a closed loop that watches how a customer's software and users behave, ranks what's worth changing, proposes the change, proves whether it worked, and carries the result into the next cycle.
View the role
Manager II, Engineering, Datadog Agent SDLC (Boston / New York)
Lead the systems, tooling, and developer experience behind the Datadog Agent itself, including software supply-chain security and evolving the SDLC as AI coding tools reshape how it gets built.
View the role
Questions or feedback on any of these? Drop a comment below.
All reactions