TEAM WORKSPACE
STAGED ALERT

Checkout fails after a bad release

Deploy an intentionally faulty checkout revision to the isolated demo service. Synthetic local requests exercise the real HTTP handler. After observed failures, an alert appears in the inbox. With the team’s staged auto-declare policy enabled, Meld creates the incident with its investigator paused. Otherwise, declare it in the inbox. Slack setup is separate and requires a verified workspace, private channel configuration, and linked responders. The service stops automatically after two hours.

Check the selected team’s Slack & staged alert settings before staging. Auto-declare is off until an owner enables it.

Checking AWS…

DEMO REPLAY

Restart the workspace demo?

This clears alerts, incidents, conversations, terminal history, and other demo history for everyone in this workspace. The exact teams with enabled Slack delivery or staged-checkout auto-declaration and their service maps stay in place; other teams and service maps are cleared. Sign-in, early-access approvals, saved source connections, and usage accounting also stay. A server backup is made first.
TEAM WORKSPACE

Create a team workspace

Name the team first. Source access and scope come next.
RELATIONSHIP REVIEW

Add relationship

Owner-reviewed links are starting context, not observed runtime traffic.
NEW INVESTIGATION

What is happening?

When selected, Meld will not start investigating until you resume the incident. Otherwise, investigation starts automatically.

Service notebooks

START WITH YOUR SERVICE

You choose the sources.
Meld puts the context together.

Connect the code and telemetry that matter. Build a draft with Meld, refine it together, and publish when it’s ready.

  1. 01
    Name your service

    Give the notebook a name and environment.

  2. 02
    Choose its sources

    Datadog and GitHub first. AWS is optional.

  3. 03
    Build, review, publish

    A useful starting point that you control.

Your notebooks

DIRECT THE SHARED INVESTIGATION

Send a direction to the Room

Review this text before sending. The team and Meld will see only the direction below; your private conversation stays here.
REVIEW BEFORE SHARING

Share a finding with the Room

Your conversation stays private. Only this claim, uncertainty, and selected terminal excerpts become visible to the team and Meld.

NEW INVESTIGATION

What’s happening?

Choose a discovered service. Meld freezes its current source context when the investigation starts.

When selected, Meld will not start investigating until you resume the incident. Otherwise, investigation starts automatically.
FULL HISTORY

Investigation record

Every observation, direction, review and decision.

Recovery criteria

Applies to this incident and future incidents for the same team, service and environment. Checks use bounded samples, not all traffic.

Record an external change

Meld does not apply changes. Record what your team did and when, so recovery checks use evidence after the change.
STEP 1 OF 4

Create a service notebook

Start with the service. You’ll choose its sources next.
ATTACH A SOURCE
WORKSPACE CONNECTION

Connect Datadog

Connect your organization here. You’ll select the service and environment in each notebook.

Create the keys in Datadog Organization Settings. Use an application key with only the read permissions Meld needs for logs, APM and metrics, preferably owned by a service account. Meld checks actual access when you attach a service.

This local prototype stores credentials in a private server-side file. Replacing this workspace connection affects every attached notebook. It does not change your deployed telemetry agents.
READY FOR INVESTIGATIONS

Publish this notebook?

New investigations can start from this version. Review known gaps so responders understand its limits. Gaps do not all need to be closed before publishing. Published text stays fixed; later changes use a new revision.
PROPOSED BY MELD

Review changes

Before

Proposed
SERVICE MAP · OWNER REVIEW

What should Meld know before the first check?

Connected source pointers are shown automatically. Add only durable service meaning; Meld verifies live state during each incident. Leave unknowns blank and record important gaps.
INVESTIGATE TOGETHER

Share with Meld

The command, output and connected reads will become cited evidence in the conversation. Your note directs Meld’s next step.



TERMINAL REFERENCE
Select output


Connect Datadog alerts
Create an authenticated endpoint for your Datadog organization.