Greptile
Also known as: Greptile Agent, TREX, greptileai, Tabnam, Inc.
AI code review agent that indexes your whole repository as a graph and keeps a self updating Knowledge Base of it, runs a swarm of parallel hypothesis scoped agents over each PR, with TREX (beta) starting the app in a sandbox and testing it, auto approve under a risk ceiling, and full self hosting including air gapped with customer supplied LLM providers.
Greptile is an AI code review agent that reviews each pull request against the whole codebase, not just the lines that changed. When a pull request opens, Greptile reviews it against a continuously updated graph of the entire repository that maps how functions, classes, files and directories connect, and against a Knowledge Base of generated documentation covering how each part of the codebase works, how it connects and which past bugs and risks it carries, updated in real time and editable from the dashboard. That lets it catch problems outside the diff, such as a change that breaks callers in files you did not touch.
Repo Clusters go further, giving reviews read only access to up to seven related repositories, so a change to a shared library is reviewed against the services that depend on it. Greptile was founded in 2023 by Daksh Gupta, Soohoon Choi and Vaishant Kameswaran and went through Y Combinator in W2024. It raised a $25M Series A led by Benchmark and a $50M Series B led by Menlo Ventures on 19 October 2026 at a $750M valuation, and it describes itself as the code validator for thousands of engineering teams including Nvidia, Coinbase, Decagon, Vanta, Stripe, Amazon and Brex, reviewing more than one percent of the pull requests opened on GitHub each day.
The v3 rewrite made the review engine agentic, and v4 raised addressed comments per pull request by 74 percent. The v5 engine replaced the single loop with a swarm of narrowly scoped agents running in parallel, each exploring one hypothesis for a potential bug. In Greptile's production A/B tests, v5 cut median review time from just over five minutes to under two and a half. It also raised the share of comments authors act on from 52 to 66 percent. Each comment carries a confidence score and a P0 to P2 severity badge, a redesigned summary leads with a confidence score and a findings list sorted by severity, and Greptile replies in the thread to follow up questions and pushback.
Two companion capabilities sit alongside review. TREX, in beta, works out which user flows a change affects, starts the app in a sandbox built from the repository's environment and uses it the way a person would, clicking and typing through each flow, writes and runs targeted tests, and posts what it tested and what broke with logs, traces and a video of every flow. A security reviewer, on by default, chains Opengrep pattern scanning, dependency analysis against known vulnerability databases, and AI exploit detection for chained issues that static tools miss.
Greptile's Memory and Learning system adapts from team pull request comments, replies to Greptile and thumbs up and down reactions. It also tracks whether comments were addressed between the first and last commit, and turns repeated patterns into suggested custom rules. Teams also write rules explicitly, as structured JSON or plain markdown in a versioned .greptile/ folder with cascading inheritance from parent to child directories, scoped by file pattern with a severity per rule.
Auto approve lets Greptile approve only pull requests that receive a clean five out of five review, under a ceiling the customer sets on the highest risk it may approve, with TREX results optionally counted. Exclusions work by label, branch, file and author, and a hard never approve list covers authentication, secrets, billing, database migrations, infrastructure, CI and public APIs.
Greptile positions itself as an independent validation layer that works with any coding agent. It reviews on GitHub Cloud, GitLab.com, Bitbucket Cloud and Cursor Origin (beta) on every plan, with GitHub Enterprise Server, GitHub Enterprise Cloud with data residency, self managed GitLab, Bitbucket Data Center and Gitea on Enterprise and Perforce on premises. It draws context from Jira, Linear, Notion, Google Drive, Sentry and Datadog and delivers findings into Slack.
A hosted MCP server works in Claude Code, Claude Desktop, Cursor and VS Code, the Knowledge Base is served to coding agents over MCP, and a plugin ships in Anthropic's official marketplace. There is also a REST API, an npm CLI with JSON and agent output modes, and a Zapier integration. Fix with your Agent routes findings straight into Claude Code, Codex, Conductor, Cursor or Devin, and Greplooping keeps an agent fixing until the review comes back clean. A Partner Program supplies review context maintained by partners for OpenAI, Vercel, Stripe, PostHog, Datadog, WorkOS, Mintlify and Braintrust.
The whole service can be self hosted, with Docker Compose on a single VM for up to 100 developers or Kubernetes above that. Greptile provides an AWS Terraform module, manual setup for GCP, Azure or on premises, a migration path between the two setups and a public deployment repository.
It can run air gapped with customer supplied LLM providers (OpenAI, Anthropic, AWS Bedrock, Azure OpenAI or GCP Vertex AI), so code and inference never leave the network. Greptile Cloud runs on AWS.
Greptile is SOC 2 Type II compliant with SAML SSO, role based access control and audit logs on Enterprise, holds zero data retention agreements with its LLM providers, and may train on de identified data unless an admin turns that off; self hosted deployments never send data for training.
Pricing is credit based with no seat fee. Pro sells prepaid credits at $1 each, shared across unlimited developers, and a review draws 1, 3 or 10 credits by depth, with each TREX run adding 3. Flex Usage Limits let teams cap that spend. A free Starter tier covers one active developer with 50 credits a month, qualified non commercial open source projects use Greptile free, early stage startups get 500 credits a month for six months, and Enterprise is custom. The legal entity is Tabnam, Inc.
From our AI code reviewers video
AI Code Review Tools: Which Ones Run the Code They Review?
Our video on the 7 AI code reviewers in the index names what each one does best. For Greptile, that is testing the change, not just reading it.
The player starts at Greptile's part, 4:04 in. Recorded in October 2026; the profile here is live and updates as research is verified.
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, raised a $25 million Series A led by Benchmark in September 2025, and on 19 October 2026 announced a $50 million Series B led by Menlo Ventures with Jump Trading, Wing, Operator Collective, Benchmark, YC, Browder Capital and Initialized participating, valuing the company at $750 million, with Deedy Das joining the board. It describes itself as the code validator for thousands of engineering teams, naming Nvidia, Coinbase, Decagon and Vanta, and earlier customers including Stripe, Amazon, Brex, Whoop, Substack, PostHog and Raycast.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Greptile reviews pull requests on GitHub Cloud, GitLab.com, Bitbucket Cloud and Cursor Origin (beta) on every plan, and on GitHub Enterprise Server, GitHub Enterprise Cloud with data residency, self managed GitLab, Bitbucket Data Center and Gitea on Enterprise, with Perforce through P4 Code Review on premises. It pulls context from Jira, Linear, Notion, Google Drive, Sentry and Datadog and delivers findings to Slack and the IDE. Fix with your Agent routes a comment with file, line numbers and suggested fix straight to Claude Code, OpenAI Codex, Conductor, Cursor or Devin via a local bridge CLI, with a Fix All control and Greplooping that keeps the agent fixing until the review is clean. A hosted MCP server at api.greptile.com/mcp works in Claude Code, Claude Desktop, Cursor and VS Code, the Knowledge Base MCP serves codebase documentation to coding agents, and a Greptile plugin ships in Anthropic's official Claude Code marketplace. A REST API exposes Index Repository, Query Repos, Search Repos and Get Repository Info, plus webhooks and Zapier. An npm CLI runs reviews locally with JSON and agent output modes. The Partner Program adds maintained review context for OpenAI, Vercel, Stripe, PostHog, Datadog, WorkOS, Mintlify and Braintrust.
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
Agentic Index coverage score
12.5 / 14 capabilities · 89%
| Integrations & Tool Calling | Full |
|---|---|
|
Reviews run on GitHub Cloud, GitLab.com, Bitbucket Cloud and Cursor Origin (beta) on every plan, with GitHub Enterprise Server, GitHub Enterprise Cloud with data residency, self managed GitLab, Bitbucket Data Center and Gitea on Enterprise and Perforce through P4 Code Review on premises. Context comes from Jira, Linear, Notion, Google Drive, Sentry and Datadog, and findings are delivered into Slack and the IDE. Fix with your Agent sends a comment with file, line numbers and suggested fix straight to Claude Code, OpenAI Codex, Conductor, Cursor or Devin through a local bridge CLI, a Fix All control sends every issue at once, Greplooping lets an agent keep fixing until the review is clean, and comments are auto marked as addressed when a commit touches the flagged files. A hosted MCP server, REST API, npm CLI and Zapier integration extend the reach, and the Partner Program adds maintained context for OpenAI, Vercel, Stripe, PostHog, Datadog, WorkOS, Mintlify and Braintrust. Sourcegreptile.com/docs/integrations/code-providers, docs/integrations/fix-with-your-agent and greptile.com/independenceread 2026-10-28 |
|
| Workflow Orchestration | Full |
|
The v5 review engine runs a swarm of narrowly scoped agents in parallel, each exploring one hypothesis for a potential bug, and aggregates their results into a single review. Greptile reports median review time falling from 5:04 to 2:25 and author addressed comments rising from 52 to 66 percent in production A/B tests. The agent looks beyond the diff for similar logic, traces dependencies, reads git history and runs iterative passes that challenge their own hypotheses, with the maximum number of agentic turns configurable per review. TREX runs as a companion agent that writes and executes tests, the security reviewer chains Opengrep pattern scanning, dependency analysis and AI exploit detection, and Greplooping hands findings to a coding agent repeatedly until the review comes back clean. Sourcegreptile.com/blog/greptile-v5, greptile.com/changelog 2026-08-05 and greptile.com/independenceread 2026-10-28 |
|
| Knowledge Grounding & RAG | Full |
|
A graph index of the entire repository maps how functions, classes, files and directories connect and updates continuously, so reviews can reason about ripple effects beyond the diff, and a Knowledge Base generates and maintains documentation of how each part of the codebase works, how it connects and which past bugs and risks it carries, updated in real time as code changes, editable from the dashboard and served to coding agents over MCP. Repo Clusters and a context.repos field give read only access to up to seven related repositories per review, with clusters suggested from shared contributors over the previous 90 days. A files.json file points reviews at existing schemas, API specs and architecture docs in the repository, and context also comes from Jira, Linear, Notion, Google Drive, Sentry and Datadog. The Partner Program supplies partner maintained documentation for supported third party APIs, and imported AI rules files, including CLAUDE.md and Cursor rules, become org wide context. Sourcegreptile.com/knowledge-base, docs/code-review/cross-repo-context and greptile.com/pricing; greptile.com/docs/code-review/cross-repo-contextread 2026-10-28 |
|
| Human Oversight & Guardrails | Partial |
|
Greptile posts its reviews and, where enabled, approves pull requests on its own after a clean five out of five review at or below a risk cap the customer sets, with exclusion filters by label, branch, file and author. Authentication, secrets, billing, migrations, infrastructure, CI and public APIs form the critical risk tier, so they are held back only when the cap sits below critical or the paths are excluded; with no cap set the page says it can approve any clean review, and an option counts TREX results in the decision. Custom rules shape what it flags. These controls constrain the agent, but no step where a person approves the agent's action before it lands is described. Sourcegreptile.com/docs/code-review/auto-approve-prs, docs/trex/overview and greptile.com/changelog 2026-06-26; greptile.com/docs/code-review/first-pr-reviewread 2026-10-28 |
|
| Security, Identity & Governance | Full |
|
Greptile is SOC 2 Type II compliant, with the report, penetration test results and subprocessor list in its Trust Center at trust.greptile.com, and it holds zero data retention agreements with its LLM providers. Single sign on with SAML and role based access control come with Enterprise, along with audit logs. Greptile Cloud runs on AWS with customer code cached on an encrypted filesystem until access is revoked, an admin can delete all customer data within 24 hours, and de identified data may be used to train Greptile's models unless an admin turns that off, which self hosted deployments never do. The whole service can also self host into a customer VPC or an air gapped environment with customer supplied LLM providers. Sourcegreptile.com/security (updated 18 October 2026) and greptile.com/pricingread 2026-10-28 |
|
| Observability & Auditability | Full |
|
Each TREX flow keeps a step by step record of what the agent did and a captioned video of the run, posted to the pull request with live progress while it runs and retained for 30 days to anyone with read access to the repository. The analytics dashboard tracks reviews, addressed rate, bugs caught and merge times with export, and Enterprise adds audit logs. The review agent itself keeps outcomes rather than a step trace. Sourcegreptile.com/docs/trex/overview, docs/trex/results and docs/analytics; greptile.com/docs/trex/resultsread 2026-10-28 |
|
| Memory & State Persistence | Partial |
|
Memory and Learning picks up team preferences from PR comments, replies to Greptile, reactions and whether comments were addressed between the first and last commit, suppresses comment types a team ignores, and turns repeated patterns into suggested rules. Its scope, lifetime and any way to review or delete what was learned are not stated. Sourcegreptile.com/docs/how-greptile-works/memory-and-learningread 2026-10-28 |
|
| Deployment & Data Residency | Full |
|
The entire service can be self hosted, including in air gapped environments, using Docker Compose on a single Linux VM for up to 100 developers or Kubernetes with Helm for 100 or more with horizontal scaling and high availability. Sizing requirements are published, and parallel deployments give a migration path between the two. An AWS Terraform module provisions VPC, EC2, RDS PostgreSQL and ElastiCache Redis and bootstraps the application, and manual setup covers GCP, Azure and on premises. Self hosted customers bring their own LLM provider by base URL and keys, choosing among OpenAI, Anthropic, AWS Bedrock, Azure OpenAI and GCP Vertex AI, so inference stays on customer infrastructure; the deployment repository is public at github.com/greptileai/akupara and self hosting requires a license. Greptile Cloud runs on AWS, and Enterprise adds GitHub Enterprise Cloud with data residency and Perforce for on premises deployments. Sourcegreptile.com/docs/deployment-options, greptile.com/security and docs/integrations/code-providersread 2026-10-28 |
|
| Prebuilt Agents, Templates & Packs | Partial |
|
The Partner Program supplies vendor curated review context for third party APIs, so a pull request that touches a supported partner API, SDK, config or integration is reviewed against partner maintained documentation, implementation rules and common failure patterns, with comments marked by partner logo and linked to the relevant docs. It is on by default for all users, with featured partners including OpenAI, Vercel, Stripe, PostHog, Datadog, WorkOS, Mintlify and Braintrust. Greptile also ships two named agents, the code reviewer and TREX, and imports existing AI rules files, including CLAUDE.md, AGENTS.md and Cursor rules, as org wide context. There is no browsable catalog, marketplace or gallery of installable agents or templates. Sourcegreptile.com/partners and greptile.com/pricingread 2026-10-28 |
|
| Triggers & Channel Coverage | Full |
|
Reviews start automatically when a pull request opens in GitHub, GitLab, Bitbucket or Cursor Origin and run again on later commits under configured review triggers. TREX runs on every reviewed pull request or only on those matching branch, label, author or path rules, and skips changes with nothing to test. Reviews can also be started with @greptileai, a re trigger control, a reply in the thread, @greptileai trex, or the CLI on a local branch. Sourcegreptile.com/docs/quickstart, docs/trex/overview and greptile.com/cliread 2026-10-28 |
|
| Model Flexibility & Routing | Full |
|
Review settings let the customer choose which AI model powers reviews and set the maximum number of agentic turns per review, balancing speed, depth and cost. Model Inversion, marked experimental, detects from commit trails, branch prefixes and PR titles when a coding agent wrote a pull request and routes the review to a different model, so Claude authored code is reviewed by GPT and GPT authored code by Claude. The cloud service combines frontier models from OpenAI and Anthropic, and self hosted deployments bring their own provider by base URL and keys, with OpenAI, Anthropic, AWS Bedrock, Azure OpenAI and GCP Vertex AI supported, so models run entirely on customer infrastructure. Sourcegreptile.com/changelog entries of 2025-09-26 and 2026-07-22, greptile.com/independence and docs/deployment-optionsread 2026-10-28 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
A REST API exposes Index Repository, Query Repos, Search Repos and Get Repository Info, alongside webhooks, with API keys issued from app.greptile.com/settings/api. A hosted MCP server at api.greptile.com/mcp, configured by bearer token in Claude Code, Claude Desktop, Cursor and VS Code, lets assistants trigger and re run reviews, inspect results, manage custom context and update repository rules, and the Knowledge Base MCP on Pro serves the maintained codebase documentation to coding agents. An npm CLI runs reviews locally, with --json for machine readable output and --agent for agent workflows and scripts, and greptile onboard sets up an organization from the terminal. A Greptile plugin ships in Anthropic's official Claude Code marketplace, and a Zapier integration is published. Sourcegreptile.com/docs/mcp/setup, greptile.com/knowledge-base, greptile.com/pricing and greptile.com/changelogread 2026-10-28 |
|
| Testing, Debugging & Optimization | Full |
|
Every pull request is reviewed for logic, security and cross file bugs, with confidence scores and P0 to P2 severity. TREX, in beta, works out which user flows a change affects, starts the app in a sandbox built from an environment the customer configures, exercises those flows like a user and runs targeted tests, posting what it tested and what broke with logs, traces and a video, so the tests act as a quality gate before release and can feed the auto approval decision. Sourcegreptile.com/trex, docs/trex/overview, docs/trex/environments and greptile.com/changelog 2026-06-15read 2026-10-28 |
|
| Browser & Computer Use | Full |
|
TREX, in beta, starts the customer's app in a sandbox built from a configured environment and uses it the way a person would, clicking, typing and observing what is on the screen, on every reviewed pull request or on those matching rules. It records a video of every flow it runs. Each TREX run costs 3 credits on top of the review, charged only when a flow reaches a pass or fail. Sourcegreptile.com/docs/trex/overviewread 2026-10-28 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
TREX, Greptile's end to end testing agent, is now generally available. On each pull request it starts the app in a sandbox, clicks through the flows the change affects, and posts what it found with a captioned video of each flow, and the release adds environment setup with network access controls, a simulation mode for testing that setup, and custom flows.
Bears on: Browser/computer use
View sourcePricing
Starter free (1 active developer, 50 credits a month). Pro is prepaid credits at $1 each, shared across the organization, with unlimited developers and no seat fees. Enterprise is custom, and a 14 day Pro trial is offered.
Prepaid credits at $1 each, shared across the organization, with no seat fees and unlimited developers. A Base review uses 1 credit, a Plus review 3 and an Apex review 10, and a TREX run adds 3. Enterprise is a custom contract, including self hosted deployment.
Included quota
Starter is free for 1 active developer with 50 credits a month. Pro sells prepaid credits at $1 each with unlimited developers and agents and no seat fee, and includes learning and custom context, the Knowledge Base MCP, cross repo context, the team analytics dashboard, and Jira, Linear, Notion and coding agent integrations. Enterprise, on a custom contract, adds self hosted and air gapped deployment, bring your own models, SSO and SAML, RBAC, GitHub Enterprise Server and Enterprise Cloud with data residency, GitLab Self Managed, Bitbucket Data Center, Gitea and Perforce support, security and compliance reviews, and a dedicated Slack channel. Pre Series A startups under $2 million revenue get 500 free credits a month for six months.
What is public
Greptile publishes a free Starter tier for 1 active developer with 50 credits a month. Pro is sold as prepaid credits at $1 each, with unlimited developers and no seat fees, and review depths cost 1, 3 or 10 credits. Qualified non commercial open source projects use it free, early startups have a credit program, and a 14 day trial is offered. Enterprise and self hosted pricing are quoted through sales.
Billing mechanics
Prepaid credits at $1 each, shared across the organization and drawn per review by depth. No seat fee. Enterprise is a custom contract with self hosting.
Cost watchouts
Every review spends credits, and deeper review modes cost three or ten times a base review, so spend tracks pull request volume and the depth chosen. TREX runs add 3 credits each. Self hosting requires a separate license, quoted by Greptile.
Variable cost rationale
Greptile meters credits at $1 each with no seat fee. Each review draws 1, 3 or 10 credits by depth, and TREX runs add 3, so the bill grows with pull request volume and the review depth chosen. Agentic workflows that open many pull requests spend credits fast. Flex Usage Limits let a customer cap spend. The free Starter tier stops at 50 credits a month for one active developer.
Additional watchouts
Every review spends prepaid credits, so high volume or agentic workflows that open many pull requests raise cost quickly. Self serve review covers GitHub and GitLab, and Bitbucket, Gitea and Perforce support comes with Enterprise.
Overage / add-ons
There is no seat allowance to exceed: every review draws prepaid credits at $1 each, by review depth (Base 1, Plus 3, Apex 10). Flex Usage Limits let a customer cap spend.
Sales call required
No, self serve available
Free / trial
Free Starter plan for 1 active developer with 50 credits a month, free use for qualified non commercial open source projects, and a 14 day Pro trial
Lowest paid plan
Pro, prepaid credits at $1 each with no seat fee
Commercial notes
Greptile prices review by the credit rather than by the seat, which suits teams with many contributors and agents. It fits teams with large, complex codebases where cross file bugs are costly, and regulated enterprises that need air gapped self hosting with their own models. It raised a $50M Series B led by Menlo Ventures in October 2026.
Key ambiguities
How re reviews are counted is not spelled out, and self hosted and enterprise pricing is quoted rather than published. The TREX run price (3 credits, charged only when a flow reaches pass or fail) sits in the TREX docs rather than on the pricing page.
Cancellation / refund
Pro credits are prepaid. Enterprise is a custom annual or multi year contract.
Support SLA / resale
SOC 2 Type II across all tiers, and Enterprise adds a dedicated Slack support channel, SSO and SAML, audit logs and self hosted deployment. No reseller or white label program is published.
Missing data
Enterprise and self hosted pricing are custom and not published.
Related vendors
- Cognition — Maker of Devin, an autonomous AI software engineer
- 10Web — Agentic website platform whose AI agents build, host and manage…
- AgentUI — Managed AI app builder for operations teams that generates and hosts…
- Aider — Open source, model agnostic terminal coding agent that edits your…
- Amp — Amp is a coding agent from Amp Frontier Corporation that runs each…
- Anthropic Claude Code — Anthropic's agentic coding system across terminal, desktop, IDE, web…
Alternatives to Greptile
The closest documented capability profiles to Greptile among coding agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Zencoder13.5 / 14Fuller documented coverage on Human Oversight & Guardrails and Prebuilt Agents, Templates & Packs
- Baz13.0 / 14Fuller documented coverage on Human Oversight & Guardrails and Prebuilt Agents, Templates & PacksGreptile vs Baz →
- Factory13.0 / 14Fuller documented coverage on Human Oversight & Guardrails and Prebuilt Agents, Templates & Packs
- Roomote13.0 / 14Fuller documented coverage on Memory & State Persistence and Prebuilt Agents, Templates & Packs
- Tembo12.0 / 14Fuller documented coverage on Prebuilt Agents, Templates & Packs
- Warp13.0 / 14Fuller documented coverage on Human Oversight & Guardrails and Prebuilt Agents, Templates & Packs
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded