Graphite

Graphite is not just an "AI review app for code." It is a developer workflow layer around pull requests.

Its product combines:
- AI code review on PRs
- PR chat for code context, CI debugging, and suggested fixes
- stacked PR workflows
- merge queue orchestration
- review inbox and team review UX
- CLI and VS Code tooling tied to GitHub

So the real business is code review and PR operations for engineering teams, with AI as one feature inside a broader collaboration and shipping workflow.

https://graphite.com

72Getting Clauded

Current verdict

Anthropic is in Graphite's lane already, especially on AI review, CI handling, and PR assistance. Graphite is not dead because it owns more of the team workflow than Claude does today, but the most marketable part of its pitch is absolutely under attack.

Assessment

Graphite is exposed because Anthropic has moved beyond generic coding help into local code review, PR handling, CI failure resolution, long-running coding agents, and codebase-scale reasoning.

That hits the flashy part of Graphite's value prop directly: AI reviews and PR collaboration.

What saves Graphite, for now, is that it appears to be a purpose-built review system for teams, not just a model. Stacked PRs, merge queues, review inboxes, GitHub-native workflow design, and enterprise developer habits create real switching friction.

Still: if your homepage leads with "AI code review platform," and Claude can already review code, comment inline, handle PRs, and fix CI in the background, you're not sleeping well.

Biggest historical hit

The clearest hit is Claude Code on desktop can now preview running apps, review code, and handle CI failures and PRs in the background.

That announcement lands squarely on Graphite's AI review and PR workflow story. It does not fully replace Graphite's stack-aware review system, but it attacks the exact surface area Graphite uses to look modern and differentiated.

What still protects them

Graphite still has some real protection if it keeps leaning into workflow product, not just model output.

Potential moats:
- Stacked PRs are a concrete workflow primitive, not just an LLM trick.
- Merge queue + inbox + PR page form a system of record for review operations.
- GitHub-native team coordination is sticky once embedded in engineering process.
- Existing adoption at recognizable engineering orgs suggests implementation and habit inertia.

What does not protect them:
- generic AI review claims
- "chat with your PR" positioning
- bug-catching and suggested-fix marketing without deeper workflow lock-in

Brutal version: if Graphite is a better-shaped wrapper around Claude-grade code review, it's vulnerable. If it's the operating system for how teams sequence, review, and land code, it still matters.

Signals

AI code review on pull requestsInline review comments before human reviewCI failure diagnosis and fixesPR handling and merge assistanceCodebase-scale software engineering reasoningAgentic coding workflows across tools

Why this is in the blast radius

Local code review: Claude leaves inline comments on bugs and issues before full code review

X / @claudeai · 2026-02-20

Inside blast radius

This is a direct strike on Graphite's AI review feature.

Graphite explicitly sells high-signal AI reviews on every PR. Claude announcing local code review with inline bug comments means Anthropic now covers the same immediate developer job: catch issues before human reviewers step in.

Claude Code on desktop can preview running apps, review code, and handle CI failures and PRs in the background

X / @claudeai · 2026-02-20

Inside blast radius

This overlaps with multiple Graphite features at once:
- code review
n- CI failure help
- PR workflow support
- background task execution

Graphite's site says its chat helps with code context and CI failures right from the PR page. Anthropic is clearly building toward the same workflow, just from the Claude side rather than a dedicated PR product.

Introducing Claude Opus 4.7

Anthropic news · 2026-04-16

Inside blast radius

Opus 4.7 is positioned as stronger at advanced software engineering, including hard coding tasks that previously required close supervision.

For Graphite, better coding models matter because AI review quality is largely downstream of model capability. As Claude gets better at understanding repositories, instructions, and verification, the value of standalone AI review tooling gets compressed unless the workflow layer is meaningfully better.

Anthropic acquires Vercept to advance Claude's computer use capabilities

Anthropic news · 2026-02-25

Inside blast radius

Graphite is a tool embedded in live developer workflows. Vercept signals Anthropic wants Claude to operate inside applications across multi-step workflows.

That increases substitution risk over time: instead of merely suggesting review comments, Claude can navigate tools, manage PR tasks, and work across the same surfaces Graphite occupies.

Introducing Sonnet 4.6 with codebase refactoring and multi-agent workflow coordination

Anthropic news · 2026-04-11

Inside blast radius

Graphite depends on Claude-class reasoning being hard to replicate at the workflow layer. Sonnet 4.6 expanding into codebase refactoring, agent coordination, connectors, and broader availability means Anthropic keeps pushing downmarket into practical engineering assistance.

That doesn't replace Graphite's merge queue or stacked PR mechanics outright, but it does erode the distinctiveness of AI help attached to reviews.

Anthropic expands partnership with Accenture to help CIOs scale Claude Code

X / @anthropicai · 2025-12-09

Outside blast radius

This is more of a distribution and enterprise adoption signal than a direct feature overlap.

It matters because Graphite sells to engineering teams and enterprises care about standardizing developer tooling. But it does not itself announce a PR review product or merge workflow substitute, so the overlap is indirect rather than immediate.

Back to home