Product deep dive

See exactly how ReleaseRay works

From merged PR to published release note — a detailed look at every step of the automation pipeline

SOC 2

Type II aligned controls

<60s

Draft generation time

3 personas

Generated per release

How ReleaseRay Works

A detailed look at each step — from merged PR to published release note

01
Ingest

PRs flow in automatically via webhook

Connect any GitHub repo with a single webhook. Merged PRs, linked issues, labels, and commit metadata are normalized and deduplicated in real time. Nothing manual, nothing to configure per-repo.

  • GitHub webhook fires on pull_request.closed + merged
  • Extracts title, body, labels, linked issues, and commit messages
  • Deduplicates across squash-merge and rebase workflows
  • Supports tag ranges, date ranges, or PR-number ranges
ingest.pipeline
POST /api/webhooks/github
→ HMAC-SHA256 verified
→ PR metadata normalized
→ linked issues resolved
→ stored in release range
02
Classify

AI groups and categorizes every change

A multi-provider LLM layer (OpenAI or your own AWS Bedrock credentials) reads each PR's labels, title, and description to classify changes into Features, Fixes, Performance, Security, and Breaking Changes.

  • Provider-agnostic: OpenAI gpt-5-mini or AWS Bedrock Claude 4.5 Haiku
  • BYOK mode keeps prompts and responses in your AWS account
  • Breaking change detection from conventional commits and labels
  • Source content processed in-memory — never persisted
classify.pipeline
labels + title + body
→ LLM classifier (BYOK optional)
→ Features | Fixes | Perf
   Security | Breaking
→ ephemeral — zero data at rest
03
Draft

Three persona-perfect drafts in under 60 seconds

Background workers generate three distinct versions simultaneously: a technical changelog for engineers, an executive summary for internal teams, and a customer-friendly update stripped of internal jargon.

  • Engineer: full PR refs, code-level detail, breaking change migration steps
  • Internal / CSM: business impact, talking points, customer-facing summary
  • Customer: plain-language feature announcements, no internal paths or links
  • Automatic redaction of internal URLs, file paths, and sensitive tokens
draft.pipeline
BullMQ background job
→ 3 parallel LLM calls
→ Engineer draft  (markdown)
→ Internal draft  (markdown)
→ Customer draft  (markdown)
→ automatic redaction pass
04
Publish

One click or scheduled — every channel at once

Approve a draft and publish immediately, or schedule delivery for the perfect time with timezone-aware scheduling. GitHub Releases, Intercom Help Center, Slack, Teams, email, and your hosted changelog — all from one screen.

  • GitHub Releases with tag attachment and markdown body
  • Intercom Help Center article creation
  • Slack and Microsoft Teams notifications
  • Hosted changelog with RSS feed and embed widget
publish.pipeline
approve → schedule (optional)
→ idempotent publish
→ GitHub Releases
→ Intercom Help Center
→ Slack / Teams / Email
→ hosted changelog + RSS

Every capability your release workflow needs

Explore features by category — click a tab to dive deeper

Smart Classification

AI analyzes PR labels, titles, and descriptions to automatically group changes into Features, Fixes, Performance, Security, and Breaking Changes.

  • Automatic categorization
  • PR context analysis
  • Breaking change detection

Persona-Aware Drafts

Generate three versions simultaneously: technical for engineers, executive brief for internal teams, and plain language for customers.

  • Engineer-focused details
  • Internal talking points
  • Customer-friendly language

Multi-Provider AI

Provider-agnostic LLM layer with OpenAI and AWS Bedrock support. Bring your own keys for full data sovereignty.

  • OpenAI + AWS Bedrock
  • BYOK data sovereignty
  • Sub-60 second generation

Ready to automate your release notes?

Start generating persona-perfect release notes in minutes — or book a walkthrough with our team.

We value your privacy

We use cookies to enhance your experience. Essential cookies are required for the site to function. You can choose to accept all cookies or manage your preferences.