Greptile
Also known as: Greptile, greptile.com, Greptile Agent, TREX
AI code review agent that builds a graph of your repository to review pull requests with full codebase context, catching cross file bugs that diff only tools miss.
Greptile is an AI code review agent built on a simple premise: a reviewer should understand the whole codebase, not just the lines that changed. When a pull request opens, Greptile reviews it against a graph it builds of the entire repository, mapping how functions, classes, files, and directories connect, so it catches problems that diff only tools are blind to, such as a change that breaks callers in files you did not touch. The index updates continuously, and developers can also ask questions of the codebase in natural language, which helps with onboarding and understanding unfamiliar systems. Founded in 2023 by Georgia Tech graduates Daksh Gupta, Soohoon Choi, and Vaishant Kameswaran, it is used by more than two thousand companies including Stripe, Amazon, and Brex.
The review engine became agentic with the v3 rewrite in late 2025 and improved again with v4 in early 2026. Rather than a single pass over the diff, the agent runs a multi hop investigation: it expands into the codebase to find similar logic, traces dependencies, checks git history to learn why code was written a certain way, and runs iterative passes that challenge their own hypotheses before posting. Each comment carries a confidence score for quick triage, and reviews include pull request summaries, inline suggestions, and auto generated sequence diagrams showing which functions call which. A companion agent named TREX writes and runs tests for every pull request in a sandbox to surface bugs and missed edge cases, and Greptile reads a team's past comments and reactions to suppress noise over time.
Greptile reviews pull requests in GitHub and GitLab across more than thirty languages, pulls context from Jira, Notion, Google Drive, and Sentry, delivers findings into Slack, and fixes issues in the IDE. Teams describe standards in plain English or a markdown file, scope rules to file patterns, and set a minimum severity so only high impact issues surface. It exposes a public API that teams use to build their own review bots and a Model Context Protocol server that lets assistants like Cursor query its rules. For regulated buyers, Greptile can run entirely inside the customer's own cloud or an air gapped environment through Docker and Kubernetes, with the option to bring your own large language models so code and secrets never leave the network.
Greptile is SOC 2 Type II certified, supports single sign on with SAML and audit logs, and is used across defense, healthcare, and financial services. Its main tradeoffs are a higher false positive rate than quieter competitors, coverage limited to GitHub and GitLab with no Bitbucket or Azure DevOps, and a pricing model that can surprise heavy users. As of the v4 release in early 2026, pricing is thirty dollars per developer each month including fifty reviews, then one dollar per additional review, which draws criticism for penalizing high volume and agentic workflows. The Developer plan is free for qualified open source projects, pre Series A startups get fifty percent off, and Enterprise adds self hosting, single sign on, and custom terms, with a fourteen day free trial.
Vendor details
Canonical URL
https://www.greptile.com
Category
Coding agent
Subcategory
AI code review agent with whole codebase context
Funding status
Independent. Founded in 2023 by Daksh Gupta, Soohoon Choi, and Vaishant Kameswaran, all Georgia Tech graduates, Greptile went through Y Combinator in the W2024 batch and raised a 25 million dollar Series A led by Benchmark in September 2025, bringing total funding to about 30 million dollars at a 180 million dollar valuation. It reports more than two thousand customers, including Stripe, Amazon, Brex, Whoop, Substack, PostHog, and Raycast, served by a team of around twenty.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Reviews pull requests in GitHub and GitLab across more than thirty languages, and pulls context from Jira, Notion, Google Drive, and Sentry, delivers findings into Slack, and lets developers fix issues in the IDE. It exposes a public API for building review bots and a Model Context Protocol server that lets assistants such as Cursor and Devin query its code quality rules. Greptile acts by leaving inline comments, suggesting click to accept fixes, answering follow ups when tagged, and, through its TREX agent, writing and running tests.
In practice
Your bugs hide across files, not in the diff. Greptile indexes your whole repository as a graph and reviews each pull request in that context, catching cross file breaks that diff only reviewers miss.
You are in a regulated industry and cannot send code to a SaaS reviewer. Greptile runs entirely inside your own cloud or an air gapped environment with your own models, so code never leaves your network.
A new engineer needs to understand an unfamiliar service. They ask Greptile in plain English how the authentication flow works and get an answer with references to the exact files, functions, and classes.
Sources & related URLs
Related / legacy domains
Capability coverage
8.5 / 14 capabilities · 61%
| Integrations & Tool CallingReviews pull requests in GitHub and GitLab across more than thirty languages, pulls context from Jira, Notion, Google Drive, and Sentry, delivers findings into Slack, fixes issues in the IDE, and exposes an API and a Model Context Protocol server, acting by commenting, suggesting click to accept fixes, and answering follow ups, broad named integrations with real action. | Full |
|---|---|
| Workflow OrchestrationRuns an agentic review engine of parallel agents that expand beyond the diff, search the codebase for similar logic, trace dependencies, check git history, and run iterative multi hop passes that challenge their own hypotheses, and a companion agent that writes and runs tests, strong orchestration that remains scoped to autonomous review and validation rather than end to end change completion. | Partial |
| Knowledge Grounding & RAGBuilds and continuously updates a graph of the entire repository, mapping how functions, classes, files, and directories connect, and reviews every change in full codebase context so it catches cross file breaks and architectural drift that diff only tools miss, deep whole codebase understanding that is the headline product. | Full |
| Human Oversight & GuardrailsPosts advisory review comments that developers act on while humans still merge, and gives teams control through custom rules written in plain English, tightly scoped file patterns, minimum severity filters, and the ability to turn off sections, a strong configurable oversight surface rather than a hard runtime enforcement gate. | Partial |
| Security, Identity & GovernanceIs SOC 2 Type II certified with independent audits, encrypts data at rest and in transit, supports single sign on with SAML and audit logs, and can run in an air gapped environment, a comprehensive identity, governance, and residency posture used across defense, healthcare, and financial services. | Full |
| Observability & AuditabilityDelivers pull request summaries, inline comments with confidence scores for triage, auto generated sequence diagrams of call flows, and impact ranked summaries, plus rule usage tracking and enterprise audit logs, real review observability short of a comprehensive agent execution tracing and analytics suite. | Partial |
| Memory & State PersistenceLearns team preferences over time by reading past pull request comments and thumbs up or thumbs down reactions and suppressing noise, but this is feedback loop adaptation and a continuously indexed codebase counted under knowledge grounding, not a distinct persistent agent memory product. | Unable to verify |
| Deployment & Data ResidencyRuns as a cloud service or entirely inside the customer's own cloud or an air gapped environment through Docker and Kubernetes, with the option to bring your own large language models, a genuine self host and data residency capability that keeps code and secrets on the customer network. | Full |
| Prebuilt Agents, Templates & PacksShips a review agent and a test writing agent named TREX and lets teams define custom rules in plain English or a markdown file and reference pattern repositories, which is user defined configuration, but not a library of prebuilt agents, templates, or packs that users browse and remix. | Unable to verify |
| Triggers & Channel CoverageTriggers automatically the moment a pull request opens in GitHub or GitLab, rescans after updates, and works across the pull request, the IDE, Slack, and a web app where developers query the codebase, real trigger and channel coverage scoped to the review workflow. | Partial |
| Model Flexibility & RoutingManages the models for its cloud reviews without user selection, but self hosted enterprise deployments can bring their own large language models and run entirely on customer infrastructure, real model flexibility through custom model support, though not a per task multi provider routing gateway. | Partial |
| APIs, SDKs & MCP ExtensibilityExposes a public API that teams use to build their own pull request review bots and a Model Context Protocol server that lets other coding assistants such as Cursor and Devin query its code quality rules, a genuine API and MCP extensibility surface, though without a documented public SDK. | Partial |
| Testing, Debugging & OptimizationMakes bug detection the core product, catching logic, syntax, style, security, and cross file issues that humans and CI miss with a leading catch rate, and its TREX agent writes and runs tests for every pull request in a sandbox to surface bugs and missed edge cases, a dedicated testing, debugging, and quality engine. | Full |
| Browser & Computer UseRuns its TREX agent in a sandbox to write and execute tests, navigates the repository graph to investigate code across files and git history, and applies click to accept fixes, real code execution and computer use within the review workflow, though not general autonomous browser automation. | Partial |
Pricing
Free for qualified open source · Pro $30/developer/mo (50 reviews included, then $1/review) · Enterprise custom (self hosting, SSO/SAML, air gapped) · 14 day free trial
Per developer per month base subscription of thirty dollars including fifty reviews, then one dollar per additional review. Free for qualified open source projects. Enterprise is a custom annual or multi year contract, including self hosted deployment.
Included quota
Developer: free for qualified open source projects licensed MIT, Apache, or GPL. Pro ($30/developer/mo): whole codebase aware review, 50 reviews per month included then $1 per additional review, 30 plus languages, custom rules, natural language codebase queries, and click to accept inline fixes. Enterprise (custom annual or multi year): self hosted and air gapped deployment, bring your own large language models, SSO and SAML, GitHub Enterprise compatibility, audit logs, dedicated Slack support, custom DPA and terms, and a higher security tier. SOC 2 Type II applies to all tiers. Pre Series A startups under 2 million dollars revenue get 50 percent off Pro; annual contracts up to 20 percent off. 14 day free trial.
What is public
The free open source Developer plan and the Pro plan at thirty dollars per developer per month including fifty reviews with one dollar per additional review are public and exact. Enterprise terms, including self hosting, are custom and not itemized.
Billing mechanics
Thirty dollar per developer per month base including fifty reviews, then one dollar per review, billed to the account. Free for qualified open source. Enterprise is a custom annual or multi year contract with self hosting. 14 day trial, no credit card.
Cost watchouts
The fifty review monthly allowance is the trap: at roughly one to one point two reviews per pull request, heavy or agentic workflows blow past it and each extra review costs one dollar, so bills can climb to several hundred dollars per seat. Enterprise self hosting is a separate custom contract. Open source maintainers have reported unexpected charges.
Variable cost rationale
The base is a predictable thirty dollars per developer per month, but it includes only fifty reviews, after which each review costs one dollar with no cap. Greptile says fewer than ten percent of users exceed the quota, but high throughput and agentic workflows blow past it fast, with reviewers reporting bills climbing to several hundred dollars per seat at a few hundred pull requests a month. Exposure is high for modern high volume and agent driven teams and low for teams comfortably under the quota.
Additional watchouts
The one dollar per review overage above fifty can escalate cost sharply for high volume or agentic workflows. Coverage is limited to GitHub and GitLab, with no Bitbucket or Azure DevOps. Higher false positive rate than quieter competitors. No in app cancel button reported. Open source maintainers have reported being billed despite the free open source policy.
Overage / add-ons
The Pro plan includes fifty reviews per developer per month; each review beyond that costs one dollar. One standard review uses one credit and the average pull request consumes roughly one to one point two reviews. There is no cap, so high volume or agentic workflows can accrue significant overage.
Sales call required
No — self-serve available
Free / trial
Free Developer plan for qualified open source projects licensed under MIT, Apache, or GPL. All other teams get a 14 day free trial with no credit card, after which Pro is thirty dollars per developer per month.
Lowest paid plan
Pro at thirty dollars per developer per month, including fifty code reviews per month with additional reviews at one dollar each, covering whole codebase aware review across GitHub and GitLab with thirty plus languages and click to accept fixes.
Commercial notes
Positioned as the highest catch rate reviewer through whole codebase indexing, priced at a modest base with usage overage. Value is highest for teams with large complex codebases where cross file bugs are costly, and for regulated enterprises that need air gapped self hosting with their own models. The usage overage makes it expensive for high throughput and agentic teams, where flat priced competitors such as CodeRabbit or Cursor BugBot are cheaper.
Key ambiguities
The real monthly cost depends heavily on review volume because of the one dollar per review overage above fifty; for agentic or high throughput teams the effective per seat cost can be several times the thirty dollar base. Enterprise self hosting pricing is custom and not public.
Cancellation / refund
Pro is a self serve monthly subscription that can be cancelled through account settings, though reviewers note there is no in app cancel button and cancellation may require contacting support. Enterprise is a custom annual or multi year commitment. Open source maintainers have reported being billed despite the free open source policy, resolved case by case.
Support SLA / resale
SOC 2 Type II certified across all tiers with independent audits and reports available on request; encryption at rest and in transit. Enterprise adds SSO and SAML, audit logs, a dedicated Slack support channel, custom DPA and terms, and self hosted or air gapped deployment. No public reseller or white label program surfaced.
Missing data
Enterprise and self hosted pricing are custom and not public beyond a reported AWS Marketplace floor. The exact economics at high review volume depend on team specific pull request and re review patterns.
Related vendors
- Cognition — Maker of Devin, an autonomous AI software engineer
- Aider — Open-source, model-agnostic terminal coding agent that works…
- Anthropic Claude Code — Anthropic's terminal and IDE-native coding agent with sub-agent…
- Augment Code — AI coding assistant evolving into an AI-native engineering platform
- Blackbox AI — AI coding assistant for code generation and developer workflows
- Blink — Agentic AI app builder that turns a natural language prompt into a…