Greptile Review 2026: Pricing, Credits & Self-Hosting

Disclosure: CodeToolPilot may earn a commission if you buy through links on this page, at no extra cost to you. How we make money.

Last verified: October 12, 2026. This is an editorial research review — how we review. Prices and plans change often; confirm on the vendor’s site.

Many AI code reviewers look mainly at the diff. Greptile is built on the opposite idea: a reviewer is only as good as its understanding of the whole codebase. It indexes your repository into a graph of files, functions and dependencies, and uses that graph when reviewing each pull request. Its homepage describes it as “AI agents that review and test pull requests with full context of the codebase.”

Greptile is aimed at engineering teams whose bugs tend to come from how changes interact with the rest of the system, not from the changed lines alone. It is also aimed at organisations that need a self-hosted or air-gapped reviewer. In 2026 it shipped a new review engine twice (v4 in March, v5 in August), introduced TREX, an agent that runs code during review, and added review tiers that let teams choose how much effort each PR gets.

This is an editorial research review. It is based on Greptile’s documentation, pricing page, changelog and blog as of October 2026, not on hands-on benchmarking. The performance figures below are Greptile’s own and are attributed as such.

Quick verdict

Greptile is one of the most technically ambitious reviewers in the category. It combines a codebase graph, a swarm of agents that each chase one possible bug, optional runtime validation and well-documented self-hosting. The trade-offs are metered pricing, where heavy users and deep review tiers can push costs past the $30 seat, a default data-sharing setting worth checking, and no listed Azure DevOps support.

Pro: Uses whole-codebase context, including read-only access to related repositories, instead of reviewing the diff alone.
Pro: Review tiers (Base, Plus, Apex, Auto) let you spend more effort only on risky PRs.
Pro: Self-hosting with Docker Compose or Kubernetes, including air-gapped installs and your own LLM providers (Enterprise).
Pro: A free Starter plan for one developer, plus free access for qualifying non-commercial open-source projects and 50% off for early-stage startups.
Con: Each seat includes 50 credits. Apex reviews cost 10 credits, and extra credits are $1 each, counted per developer rather than pooled.
Con: Azure DevOps is not among the supported providers, and self-managed Git hosts require Enterprise.
Con: On the cloud service, an “improve Greptile” data-sharing setting is on by default, and the security page allows anonymised data to be used for model training unless you opt out.
Con: TREX is in beta and cannot yet be combined with the Plus, Apex or Auto tiers.

Visit Greptile

What is Greptile?

Greptile is an AI code review service that connects to your Git provider, builds a graph of each repository, and posts review comments on pull requests. It also learns from how your team responds. Comments carry P0, P1 or P2 severity badges, and each comment has buttons that hand the fix to a coding agent.

The engine changed a lot this year. Greptile v4 (March 2026) came with new usage-based pricing. Greptile v5 (August 2026) rebuilt the reviewer as a “swarm of agents that each explore one hypothesis for a potential bug” and is now enabled for all users. In its own A/B test on production data, Greptile reports that v5 cut median review time roughly in half (from “5:04 to 2:25”) and raised the share of its comments that authors addressed from 52% to 66%. These are vendor-reported figures.

Supported code providers on Greptile Cloud, on all plans, are GitHub.com, GitLab.com, Bitbucket Cloud and Cursor Origin (beta). With Enterprise access, Greptile also supports GitHub Enterprise Server, GitHub Enterprise Cloud with data residency, GitLab Self-Managed, Bitbucket Data Center and Gitea. Perforce P4 Code Review is supported only on self-hosted Greptile.

Key features

Codebase graph and multi-repo context

Greptile indexes files, functions and dependencies so it can judge how a change affects code outside the diff. Since April 2026, a context.repos setting gives reviews read-only access to related repositories. This is useful when a client library and a service live in separate repositories.

Review tiers

Introduced in September 2026, tiers set how much work goes into each review. Base costs 1 credit, Plus 3 credits and Apex 10 credits, and Auto lets Greptile choose per PR. Tiers can be set for the organisation, per repository or directory in .greptile/config.json, through filter rules (for example by label or file path), or for a single PR with a comment such as @greptileai review this at apex. A PR comment can raise a tier but not lower it.

TREX runtime validation

Launched in June 2026 and still labelled beta, TREX lets Greptile run code during review. It sets up a sandbox, runs the relevant code and attaches evidence such as logs, screenshots, API traces and videos of UI changes. Greptile says reviews with TREX catch 20% more bugs in its internal evaluations. A TREX review costs 3 credits.

Learning and custom rules

Greptile learns from thumbs-up and thumbs-down reactions and from replies, and can infer rules from repeated review comments. Teams can also write standards in plain English. The documentation’s own illustrative timeline suggests learning takes several weeks to settle, so expect more noise in the first month.

Agent hand-off and MCP

“Fix in Claude Code”, “Fix in Codex” and “Fix in Cursor” buttons send a single issue, or all issues, to a coding agent. An MCP server lets agents read Greptile’s review comments and context, and the docs cover Jira and Linear integrations.

Self-hosting

Enterprise customers can run Greptile in their own AWS, GCP or Azure environment with Docker Compose or Kubernetes, including fully air-gapped. It works with any OpenAI-compatible model API, custom LLM implementations or AWS Bedrock. Greptile states that with on-premises deployment, code never leaves the customer’s environment.

Who should use Greptile

Greptile suits teams with large or tightly coupled codebases, monorepos, or several interdependent repositories, where context matters more than raw comment volume. It is also a strong candidate for organisations that must self-host or run air-gapped and want to choose their own model provider, and for teams already working with Claude Code, Codex or Cursor that want one-click hand-off of fixes.

Who should skip it

Azure DevOps shops should look elsewhere. Teams that want a flat, predictable bill, or that expect a few developers or agents to generate a large share of PRs, may find per-developer credit metering awkward. Teams that cannot accept any default sharing of review activity should only proceed once they have changed the relevant settings.

Strengths and limitations

Strengths

Context beyond the diff. The codebase graph and multi-repo context target the bugs that diff-only review misses, such as broken callers, mismatched contracts and side effects in other modules.
Effort matched to risk. Review tiers and filter rules let teams send migrations, auth changes or large refactors to Apex while routine PRs stay on Base.
Deployment flexibility. Detailed self-hosting docs, air-gapped support and bring-your-own-model make Greptile realistic for defence, healthcare and finance teams. It is also one of the few reviewers that supports Gitea and Perforce.
Fast product cadence. Two major engine releases, runtime validation and review tiers within one year show active development, with changes documented in a public changelog.
Compliance basics. Greptile says it is SOC 2 Type II compliant and provides the report on request. Admin-initiated deletion hard-deletes customer data from production within 24 hours, and backups are destroyed within 30 days.

Limitations

Metered costs per developer. Fifty credits is 50 Base reviews but only 5 Apex reviews. Because overage is per author rather than pooled, one prolific developer, or a coding agent opening many PRs under one account, can drive up the bill. Set a flex limit.
Data defaults need attention. Greptile’s cloud service caches your code on its servers until you revoke access, and uses OpenAI and Anthropic for inference. The security page says Greptile may use anonymised customer data to train and improve models, with an opt-out in settings, and the “Help us improve Greptile” setting is on by default. Review both before you connect sensitive repositories.
Provider gaps. Azure DevOps is not listed as supported. GitHub Enterprise Server, GitLab Self-Managed and Bitbucket Data Center all require Enterprise.
Beta features. TREX is in beta and incompatible with higher tiers for now, and Cursor Origin support is also in beta.
Self-hosting terms. According to the self-hosting docs, self-hosted Greptile uses annual contracts with no free trial, though a refund is available within the first 30 days.
Warm-up period. The learning system needs weeks of feedback before reviews fully reflect team preferences.

Pricing (as of October 2026)

Plan	Price	What you get
Starter	Free	One active developer, 50 credits per month, unlimited repositories
Pro	$30 per seat/month	50 credits per seat, extra credits at $1 each, unlimited repositories and users, custom rules, unlimited external app connections; 14-day free trial
Enterprise	Custom	Self-hosting option, SSO/SAML, GitHub Enterprise and other self-managed providers, dedicated Slack support, custom invoicing, custom DPA and terms

A seat is any developer who received a Greptile review in the billing period. Credit costs are 1 for a Base review, 3 for Plus or TREX, and 10 for Apex. Extra usage (“flex”) is calculated per developer, and admins can cap it or set it to $0. Qualifying non-commercial open-source projects get Greptile free, and pre-Series A startups can apply for 50% off. Check Greptile’s pricing page for current terms before you subscribe.

See Greptile pricing

Prices verified on October 12, 2026 against the official Greptile pricing page. Plans change often — confirm on the vendor’s page before you buy.

Use cases

Greptile’s core idea is that a reviewer needs to understand the whole codebase, not just the diff. The scenarios below are based on Greptile’s documented features as of October 2026, and each notes the cost or maturity caveat that applies.

Catching bugs outside the diff in large or multi-repo codebases

Teams with a large monorepo, or with a service and its client library in separate repositories, often ship bugs that sit outside the changed lines, such as a broken caller or a mismatched contract. Greptile reviews each PR against a graph of files, functions and dependencies, and its context.repos setting gives reviews read-only access to related repositories. Expect more noise in the first weeks, because Greptile’s learning from reactions and replies takes time to settle.

Spending review effort where the risk is

A tech lead can route database migrations, authentication changes or large refactors to the Apex tier through filter rules on labels or file paths, while routine PRs stay on Base. Developers can also raise the tier for a single PR with a comment such as @greptileai review this at apex. Because an Apex review costs 10 credits against a Base review’s 1, and overage is counted per developer, set a flex limit before rolling this out widely.

Self-hosted review for defence, healthcare or finance

Organisations that cannot send code to a third-party cloud can run Greptile on the Enterprise plan in their own AWS, GCP or Azure environment with Docker Compose or Kubernetes, including fully air-gapped. They can use any OpenAI-compatible model API, a custom LLM implementation or AWS Bedrock, and Greptile states that code never leaves the customer’s environment with on-premises deployment. Self-hosting is also the only way to use Greptile with Perforce P4 Code Review. The trade-off is an annual contract with no free trial, although the self-hosting docs mention a refund within the first 30 days.

Handing review findings straight to a coding agent

Teams already working with Claude Code, Codex or Cursor can use the “Fix in” buttons on Greptile comments to send one issue, or all of them, to that agent. Greptile’s MCP server also lets agents read review comments and context directly. If agents open many PRs under one developer’s account, that account’s credit usage can climb quickly, so watch per-developer flex spending.

Getting runtime evidence for UI and API changes

Front-end and API teams can turn on TREX, which sets up a sandbox during review, runs the relevant code and attaches evidence such as logs, screenshots, API traces or videos of UI changes. That gives reviewers something concrete to check rather than static comments alone. TREX is still in beta, costs 3 credits per review, and cannot yet be combined with the Plus, Apex or Auto tiers.

Alternatives

CodeRabbit supports Azure DevOps along with GitHub, GitLab and Bitbucket, and bills seats only for developers who open PRs.
Qodo emphasises rules and governance, with pooled team credits and on-premises or air-gapped Enterprise deployments.
Graphite is GitHub-only, but its Team plan includes unlimited AI reviews alongside stacked PRs and a merge queue.
GitHub Copilot code review comes with paid Copilot plans and needs no extra vendor.

Final verdict

Greptile is one of the strongest choices for teams that want an AI reviewer that understands the whole codebase and are willing to manage a metered bill. Its review tiers, runtime validation and self-hosting options are real differentiators. Before you roll it out, set a flex spending limit, decide which PRs deserve Plus or Apex, and review the data-sharing settings.

Try Greptile

FAQ

Is Greptile free?

Greptile’s Starter plan is free for one active developer and includes 50 credits per month with unlimited repositories. The paid Pro plan comes with a 14-day free trial. Qualifying non-commercial open-source projects can use Greptile free, and pre-Series A startups can apply for a 50% discount.

How do Greptile credits and seats work?

A seat is any developer who received a Greptile review in the billing period, and each seat includes 50 credits. A Base review costs 1 credit, a Plus or TREX review 3 credits, and an Apex review 10 credits. Extra usage is calculated per developer rather than pooled, and admins can cap it or set it to zero. Check Greptile’s pricing page for the current rate on extra credits before you budget.

Does Greptile support Azure DevOps or GitLab?

Greptile Cloud supports GitHub.com, GitLab.com, Bitbucket Cloud and Cursor Origin (beta) on all plans. GitHub Enterprise Server, GitHub Enterprise Cloud with data residency, GitLab Self-Managed, Bitbucket Data Center and Gitea require Enterprise, and Perforce is supported only on self-hosted Greptile. Azure DevOps is not listed as supported, so Azure DevOps teams should look at CodeRabbit or Qodo instead.

Does Greptile train on my code?

On the cloud service, Greptile’s security page says it may use anonymised customer data to train and improve models, with an opt-out in settings, and the “Help us improve Greptile” setting is on by default. Greptile also caches your code on its servers until you revoke access and uses OpenAI and Anthropic for inference. With on-premises deployment, Greptile states that code never leaves your environment. In our view, admins should review both data settings before connecting sensitive repositories.

Can Greptile be self-hosted?

Yes, for Enterprise customers. Greptile can run in your own AWS, GCP or Azure environment with Docker Compose or Kubernetes, including fully air-gapped, and works with OpenAI-compatible model APIs, custom LLM implementations or AWS Bedrock. Self-hosted contracts are annual with no free trial, although the docs mention a refund within the first 30 days.

Greptile vs CodeRabbit: what is the difference?

In our assessment, Greptile’s distinguishing features are whole-codebase context, selectable review tiers, runtime validation with TREX and air-gapped self-hosting with your own models. CodeRabbit supports Azure DevOps alongside GitHub, GitLab and Bitbucket, and bills seats only for developers who open PRs. Teams that want deep context on a complex codebase may lean toward Greptile, while Azure DevOps shops or teams that want simpler seat-based billing may prefer CodeRabbit. GitHub-only teams that also want stacked PRs and a merge queue should compare Graphite as well.

Greptile Review (2026): Is Codebase-Aware AI Code Review Worth It?