Back to vendors
G

Graphite

Also known as: Graphite, graphite.dev, Diamond, Graphite Agent, Graphite Reviewer

Visit site
Coding agentacquiredVerified 2026-06-30

AI code review and pull request platform built around stacked pull requests, a stack aware merge queue, and Graphite Agent, which reviews, chats, fixes, and merges code.

Graphite is an AI powered code review and pull request platform built for the outer loop of software development, the reviewing, merging, and shipping that becomes the bottleneck as AI accelerates code generation. Its signature idea is stacked pull requests, a workflow borrowed from Meta and Google in which a large change is broken into a chain of small dependent pull requests, each reviewed and merged on its own, so a developer can keep working up the stack while earlier parts are still in review. A command line tool and a VS Code extension handle the stacking and automatically rebase the stack as changes land. Founded in 2020 by Merrill Lutsky, Greg Foster, and Tomas Reimers, Graphite was acquired by Anysphere, the maker of Cursor, in December 2025, and continues to operate as its own product.

The AI layer, first launched as Diamond and now unified with chat under the name Graphite Agent, reviews every pull request in seconds with high signal feedback rather than noise. It catches real bugs, logic errors, style inconsistencies, and security issues in the context of the whole codebase, and teams report it surfacing problems that human reviewers and CI both miss. Graphite Agent goes past commenting: developers chat with it in the pull request to ask questions, refine suggestions, and apply fixes conversationally, taking a change from feedback to commit in one flow. Teams import their own style guide and write rules in plain English, and its feedback is tuned continuously against whether developers accept or reject each comment.

Around the reviewer sits a full workflow platform. A stack aware merge queue organizes approved pull requests to keep the trunk branch green, CI optimizations integrated with Buildkite and GitHub Actions let teams selectively trigger and batch test runs, and a pull request inbox centralizes reviews. Developer productivity insights give engineering leaders analytics on throughput and cycle time, and a workflow automation engine automates parts of review and merge. Graphite plugs into GitHub repositories and pipelines, works in VS Code and the command line, and now surfaces Cursor cloud agents inside the pull request so code can be created, reviewed, and shipped without leaving the page.

Graphite is SOC 2 Type II certified, continuously penetration tested, and its AI features are opt in and do not train on customer code. Its main constraints are that it centers on GitHub and is delivered as hosted software with no self host option, it offers no user facing model choice and no public API to build on, and the stacked workflow has a learning curve. Pricing simplified in early 2026: a free Hobby tier with the stacking command line tool, the VS Code extension, and limited AI reviews, a Starter plan at twenty dollars per user each month billed annually, a most popular Team plan at forty dollars per user each month that unlocks unlimited Graphite Agent reviews and chat, the merge queue, and automations, and a custom Enterprise plan, each with a thirty day free trial and no credit card.

Vendor details

Canonical URL

https://graphite.dev

Category

Coding agent

Subcategory

AI code review and pull request platform

Funding status

Acquired. Graphite was acquired by Anysphere, the maker of Cursor, in December 2025 for a price reported to be well above its prior valuation of about 290 million dollars, and continues to operate independently with the same team, brand, and product. Founded in 2020 by Merrill Lutsky, Greg Foster, and Tomas Reimers, it had raised a 20 million dollar Series A led by Andreessen Horowitz and a 52 million dollar Series B led by Accel, with Menlo Ventures participating through the Anthology Fund. More than one hundred thousand engineers across over five hundred companies, including Asana, Ramp, and Vercel, have used it.

Company status

acquired

Use cases & customers

Primary use cases

stacked pull request workflowautomated AI code reviewstack aware merge automationdeveloper productivity analytics

Target customers

high velocity software engineering teamsteams on GitHubengineering leaders tracking productivity

Deployment options

SaaS (GitHub app)CLIVS Code extension

Integrations

Integrates deeply with GitHub across repositories, pull requests, and CI, and works through a command line tool, a VS Code extension, a web pull request page, and a pull request inbox. It optimizes CI through Buildkite and GitHub Actions, and following the Cursor acquisition it surfaces Cursor cloud agents inside the pull request. Graphite Agent acts by commenting inline, applying fixes, updating pull requests, and merging through a stack aware merge queue, and it automatically rebases stacked branches as changes land.

In practice

Your large changes sit in review too long and block the team. Graphite lets you split them into a stack of small dependent pull requests, each reviewed and merged on its own while you keep building.

AI is generating more code than your reviewers can keep up with. Graphite Agent reviews every pull request in seconds, flags real bugs and security issues in context, and lets you apply fixes in chat.

Your trunk keeps breaking and merges pile up. Graphite's stack aware merge queue sequences approved pull requests and batches CI so the main branch stays green without engineers babysitting the merge order.

Capability coverage

6.0 / 14 capabilities · 43%

Integrations & Tool CallingIntegrates deeply with GitHub across repositories, pull requests, and CI, works through a command line tool, a VS Code extension, and a pull request inbox, optimizes CI with Buildkite and GitHub Actions, and now surfaces Cursor cloud agents in the pull request, acting by commenting, applying fixes, updating pull requests, rebasing stacks, and merging through a stack aware merge queue. Full
Workflow OrchestrationOrchestrates the outer loop with a stack aware merge queue that sequences and merges approved pull requests to keep the trunk green, CI batching, automatic rebasing of stacked branches, and a review to fix to merge flow, real workflow orchestration, though it is human collaborative and rules based rather than an autonomous multi agent engine. Partial
Knowledge Grounding & RAGGrounds every review in the context of the whole codebase, delivering high signal, contextually relevant feedback that catches cross file and long standing bugs that human reviewers and CI miss, with a dedicated evaluation framework tuning relevance, a headline codebase understanding capability. Full
Human Oversight & GuardrailsIs built on the principle that AI augments rather than replaces human review, with the human reviewer accountable for what ships; Graphite Agent proposes fixes and suggestions that the developer applies and merges, and teams set plain English rules for what to flag, a strong advisory oversight surface rather than a hard runtime enforcement gate. Partial
Security, Identity & GovernanceIs SOC 2 Type II certified and continuously penetration tested, and its AI features are opt in and do not train on customer code, a solid security posture, though an enumerated identity and governance matrix with SSO, SCIM, RBAC, and audit is not documented for the code review product. Partial
Observability & AuditabilityProvides review visibility through a pull request inbox and a modern review interface, and gives engineering leaders developer productivity insights with analytics on throughput and cycle time, real observability into the review workflow short of comprehensive agent execution tracing and audit. Partial
Memory & State PersistenceTunes its reviews continuously from whether developers accept or reject each comment and learns team conventions from imported rules, but this is feedback loop adaptation grounded in review history, not a distinct persistent agent memory product. Unable to verify
Deployment & Data ResidencyDelivered as hosted software through a GitHub app with a command line tool and a VS Code extension; no self host, on premises, private instance, or data residency deployment option is documented. Unable to verify
Prebuilt Agents, Templates & PacksLets teams import a style guide and write custom rules in plain English and configure workflow automations, which are user defined configuration, but ships a single reviewer agent rather than a library of prebuilt agents, templates, or packs. Unable to verify
Triggers & Channel CoverageTriggers automatically on pull request open and update and on merge queue events, integrates with CI runs, and works across GitHub, a VS Code extension, a command line tool, and a pull request inbox, real trigger and channel coverage scoped to the review and merge workflow. Partial
Model Flexibility & RoutingUses AI models internally, including Anthropic's API for pull request summaries, and tunes them for review quality, but no user facing model choice, bring your own key, or routing gateway is documented. Unable to verify
APIs, SDKs & MCP ExtensibilityOffers a powerful command line tool and a VS Code extension as client interfaces, but no public platform API, SDK, or MCP server for building on Graphite is documented, and customization is done through in product rules and automations. Unable to verify
Testing, Debugging & OptimizationMakes bug and quality detection the core product, reviewing every pull request for real bugs, logic errors, style inconsistencies, and security vulnerabilities, turning comments into applied fixes, and self healing failing CI, a headline testing, debugging, and quality capability. Full
Browser & Computer UseReviews and navigates code to produce fixes, applies commits, automatically rebases stacked branches through its command line tool, and self heals failing CI, real code operation and computer use within the review and merge workflow, though not general autonomous browser automation. Partial

Recent platform changes

No recent material changes tracked yet.

Pricing

Free (Hobby: stacking CLI, VS Code, limited AI reviews) · Starter $20/user/mo annual · Team $40/user/mo annual (unlimited Graphite Agent + merge queue) · Enterprise custom

Per user per month subscription tiers billed annually, with a free tier and a 30 day trial. The Team tier includes unlimited AI reviews and chat with no per review metering; enterprise is custom.

Public — exactLow variable costFree tierTrial available

Included quota

Free/Hobby: stacking command line tool, VS Code extension, and limited Graphite Agent AI reviews. Starter ($20/user/mo annual): all GitHub organization repositories and team insights. Team ($40/user/mo annual, most popular): unlimited Graphite Agent AI reviews and chat, stacked pull requests, stack aware merge queue, review customizations, automations, and advanced team features. Enterprise: custom with advanced controls, security features, and support. All plans include a 30 day free trial with no credit card.

What is public

All tier names and prices are public and exact: Free Hobby, Starter at twenty dollars per user per month, Team at forty dollars per user per month billed annually, and custom Enterprise. Enterprise terms are not itemized.

Billing mechanics

Per user per month subscription billed annually across three self serve tiers plus custom Enterprise. Team is the unlimited AI tier. Free tier and a 30 day no credit card trial on all plans.

Cost watchouts

Cost scales with the number of user seats on the Team tier; unlimited AI reviews avoid per review surprises. Annual billing is required for the listed per user prices. Enterprise pricing is quote based.

Variable cost rationale

Pricing is a predictable per user monthly subscription, and the most popular Team plan includes unlimited Graphite Agent reviews and chat with no per review metering, so spend scales with seat count and tier rather than usage. The main step is choosing Team for unlimited AI. Exposure is low: bounded per seat with no usage overage.

Additional watchouts

Centered on GitHub with no support for other Git hosts. Delivered as hosted software with no self host or data residency option. No user facing model choice. The stacked pull request workflow has a learning curve, and some users dislike the diff and comment panel layout. As an Anysphere property, the longer term roadmap may fold into Cursor's platform.

Overage / add-ons

No per review or usage overage on the Team tier, which is unlimited for AI reviews and chat. Cost changes come from adding user seats or moving between tiers rather than metered usage.

Sales call required

No — self-serve available

Free / trial

Free Hobby tier includes the stacking command line tool, the VS Code extension, and a limited amount of Graphite Agent AI review. Every paid plan includes a 30 day free trial that does not require a credit card.

Lowest paid plan

Starter at twenty dollars per user per month billed annually: support for all GitHub organization repositories and team insights, on top of the free stacking command line tool and VS Code extension.

Commercial notes

Repriced in early 2026 to bring AI reviews, stacking, and the merge queue into one Team plan at forty dollars per user, positioned against Cursor BugBot at the same price. Now owned by Anysphere, so expect tighter Cursor integration over time. Value is highest for high velocity GitHub teams that want the stacked pull request workflow plus AI review in one platform.

Key ambiguities

How the product and pricing evolve under Anysphere ownership, including potential bundling with Cursor, is not yet public. The exact AI review allowance on the free tier is unspecified.

Cancellation / refund

Self serve monthly plans billed annually, with a 30 day free trial and no credit card to start. Enterprise via custom agreement. Specific cancellation and refund terms are not detailed.

Support SLA / resale

SOC 2 Type II certified with continuous penetration testing; report available from the security team. AI features are opt in and do not train on customer code. Enterprise adds advanced controls and support. No public reseller or white label program surfaced.

Missing data

Enterprise pricing is custom and not itemized. Post acquisition changes to packaging or bundling with Cursor are not yet public. Exact limits of the free tier AI reviews are not specified.

Verified 2026-06-30

Contact us

Found a vendor we missed? Have feedback on the index? We'd love to hear from you.

Agentic AI Index

A directory and comparison resource for AI agent platforms, autonomous workflow tools, and enterprise agentic automation products.

© 2026 Agentic AI Index

3801 N Capital of Texas Hwy, Ste E240 · Austin, TX 78746

Researched from public vendor sources. See Methodology.