DebugBundle

When a bundle has sufficient request context, DebugBundle automatically generates reproduction artifacts — ready-to-run commands and structured specifications that recreate the failing request.

Reproduction Artifacts

Each reproduction includes three formats:

cURL

A complete, ready-to-run cURL command:

curl -X POST 'https://api.example.com/checkout' \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json' \
  --data-raw '{"cart_id":"cart_01H...","step":"payment"}'

HTTPie

A complete HTTPie command:

printf '%s' '{"cart_id":"cart_01H...","step":"payment"}' | \
  http POST 'https://api.example.com/checkout' \
  'Content-Type:application/json' \
  'Accept:application/json'

JSON Spec

A structured, machine-readable specification:

{
  "method": "POST",
  "url": "https://api.example.com/checkout",
  "headers": {
    "Content-Type": "application/json",
    "Accept": "application/json"
  },
  "query": {},
  "body": {
    "cart_id": "cart_01H...",
    "step": "payment"
  }
}

The JSON spec is designed for programmatic use — agent tools, test frameworks, and CI pipelines can consume it directly.

Reproduction Confidence

Each reproduction includes a confidence score (0–1) and a reason:

Field	Description
`possible`	`true` if reproduction can be generated; `false` otherwise.
`confidence`	Numeric confidence score (0.0 to 1.0).
`reason`	Why reproduction is (or isn't) possible: `request_context_available`, `insufficient_context`, etc.

Feasibility by Scenario

Different types of bugs have different reproduction feasibility:

Scenario	Feasibility	Notes
Standard HTTP bugs	High	Direct request replay usually reproduces the issue.
Frontend interaction + failing request	Medium-high	Request context is available; browser state may differ.
Background jobs	Medium	Job payload may be available, but timing and queue state matter.
External outage timing	Low-medium	Third-party state can't be replayed.
Race conditions	Low	Timing-dependent; single request replay may not trigger.

What Gets Included

The reproduction engine builds artifacts from the request context captured by the SDK:

Data Source	Used For
`context.request.method`	HTTP method in the command.
`context.request.path`	URL path.
`context.request.query`	Query string parameters.
`context.request.headers`	Request headers (sensitive values are redacted).
`context.request.body`	Request body.
`context.response`	Not included in reproduction (used for bundle context only).

Sensitive data is redacted before reproduction artifacts are generated. Authorization headers, cookies, and other sensitive values are replaced with [REDACTED]. This ensures reproduction commands are safe to share in issue trackers, chat, and CI logs.

Accessing Reproductions

API

curl https://api.debugbundle.com/v1/incidents/inc_01H.../reproduction \
  -H "Authorization: Bearer dbundle_member_a1b2c3d4..."

CLI

debugbundle incidents show inc_01H... --reproduction

MCP

get_reproduction(incident_id: "inc_01H...")

When Reproduction Isn't Possible

Reproduction artifacts require request context. They won't be generated when:

The error occurred outside an HTTP request (background jobs, cron tasks).
The SDK didn't capture request context (configuration, SDK limitations).
The error is in a frontend-only path with no triggering network request.
The event type is log_event, deploy_metadata, or probe_event.

In these cases, reproduction.possible is false and reproduction.reason explains why.

Using Reproductions in CI

The JSON spec format integrates well with test frameworks:

import spec from "./reproduction.json";

test("reproduce incident inc_01H...", async () => {
  const response = await fetch(spec.url, {
    method: spec.method,
    headers: spec.headers,
    body: JSON.stringify(spec.body),
  });

  // Verify the fix: this should no longer return 500
  expect(response.status).not.toBe(500);
});

Next Steps

Incidents — Incident lifecycle and grouping
Bundle Schema — Full bundle JSON structure
MCP Workflows — Agent-driven investigation patterns

Reproduction