Qodo
Also known as: CodiumAI, PR-Agent
An AI code review and governance platform, formerly CodiumAI, with parallel specialist review agents and a judge agent, a Context Engine that indexes the codebase and PR history, a Software Map of repository dependencies, Rule Miner that turns accepted review patterns into rules, PR Triage, and an Agentic Toolbox that brings Qodo's skills into Claude Code, Codex and Kiro.
Qodo, formerly CodiumAI, is an AI code review and governance platform. It sits outside the code generation loop and checks changes whichever assistant wrote them, reviewing pull requests on GitHub, GitLab, Bitbucket and Azure DevOps, on Gerrit for Enterprise customers, and reviewing changes before a pull request through the Agentic Toolbox inside a coding agent. Qodo 3.0, released 1 October 2026, positions the platform as quality control for what Qodo calls the agentic software factory.
Review is split across specialist agents for logic and bugs, standards, architecture and risk that run in parallel, and a judge agent consolidates their findings, removes duplicates and filters by confidence. A finding recommendation agent ranks each issue against how the team handled similar findings before, so feedback the team usually ignores is marked as such.
Spec and Design review agents, in research preview, check an implementation against linked specification documents and Figma designs, a Remediation agent, also in research preview, applies fixes above a severity threshold and opens them as a separate pull request, and review effort mode matches review depth to the risk of each change. PR Triage, in research preview, puts open pull requests from every repository and provider into one queue ranked by SLA, priority and blast radius.
The Context Engine is the knowledge layer behind every agent. It indexes the codebase to reason across modules and services, PR history indexing keeps a running record of past review decisions, and Software Map, in beta, keeps a live map of how repositories, services and contracts connect and the blast radius of each change.
Cross repo review flags a change in a shared library against the services that depend on it, even when they sit on different Git providers. Rule Miner mines accepted review patterns into rules scoped to an organization, repository or path, which activate directly or wait for an admin to approve, and fade when their suggestions keep being dismissed, and skill files found across repositories are governed the same way.
The Agentic Toolbox, generally available since September 2026, packages Qodo's skills (Codebase Wisdom, Get Rules, Review, Review Resolver, Manage Standards) for other coding agents, as verified plugins on the Claude Code, Codex and Kiro marketplaces and over MCP or a CLI, so an agent can review and fix its own work before opening a pull request. The IDE plugins, formerly Qodo Gen, retire on 30 October 2026 in favor of the Toolbox. Deployment runs from multi tenant and single tenant SaaS to on premises and air gapped.
Qodo says it is SOC 2 Type II certified, analyzes code and discards it under zero data retention, and offers SSO, audit logs and bring your own key for OpenAI, Anthropic, Azure OpenAI or self hosted models on Enterprise. Pricing is pooled credits at $0.012 each on Pro Team, with a 14 day trial and Enterprise custom. The original PR-Agent project continues as open source under community governance, separate from the commercial platform.
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 Qodo, that is turning your own review history into enforced rules.
The player starts at Qodo's part, 6:21 in. Recorded in October 2026; the profile here is live and updates as research is verified.
Vendor details
Canonical URL
https://www.qodo.ai/
Category
Coding agent
Funding status
Independent. Founded 2022 as CodiumAI, rebranded to Qodo. Early investors include TLV Partners. Built its reputation on the open source PR-Agent, which it donated to community governance under Apache 2.0 in early 2026, now maintained independently as a legacy project distinct from the commercial platform.
Company status
independent
Use cases & customers
Target customers
Deployment options
Integrations
Git platform integrations cover GitHub cloud and Enterprise Server, GitLab cloud and self managed, Bitbucket Cloud and Data Center and Azure DevOps for automated pull request review, with Gerrit on Enterprise and the open source PR-Agent extending to CodeCommit and Gitea. The Agentic Toolbox ships as verified plugins for Claude Code, Codex and Kiro and connects over MCP or a CLI, so coding agents can fetch rules, review local changes and resolve findings before a pull request exists; the IDE plugins for VS Code and JetBrains, formerly Qodo Gen, retire on 30 October 2026 in its favor. People chat with Qodo in a pull request by addressing it as qodo, and Enterprise bring your own key lets the customer supply its own OpenAI, Anthropic, Azure OpenAI or self hosted models.
In practice
Your team writes more code with Copilot and Claude Code than anyone can review. Qodo reviews every pull request with parallel specialist agents and a judge that filters to the findings worth acting on, independent of whichever tool wrote the code.
Review comments keep repeating the same conventions. Rule Miner turns the patterns your team keeps accepting into rules scoped to the repository or path, so they are enforced on the next review without anyone writing them down.
You're in a regulated industry and code can't leave your network. Qodo deploys single-tenant, on-premises or air-gapped, with bring your own key on Enterprise.
Sources & related URLs
Related / legacy domains
Agentic Index coverage score
12.0 / 14 capabilities · 86%
| Integrations & Tool Calling | Full |
|---|---|
|
Git platform integrations cover GitHub cloud and Enterprise Server, GitLab cloud and self managed, Bitbucket Cloud and Data Center, Azure DevOps, and Gerrit on Enterprise across multi tenant, single tenant and on premises deployments, with the open source PR-Agent adding CodeCommit and Gitea. Before a pull request exists, the Agentic Toolbox installs as verified plugins on the Claude Code, Codex and Kiro marketplaces or over MCP and a CLI, so coding agents can fetch rules, review local changes and resolve findings; the IDE plugins, formerly Qodo Gen, retire on 30 October 2026 in favor of the Toolbox according to their marketplace changelog, and the docs no longer carry an IDE section although the homepage FAQ still lists VS Code, JetBrains and Visual Studio. Chat with Qodo answers in the pull request when addressed as qodo, and reviews draw ticket context from linked issues and Azure Boards work items. Sourceqodo.ai, docs.qodo.ai/llms.txt, docs.qodo.ai/whats-new (Qodo 2.9, 2.10 and 3.0) and docs.pr-agent.airead 2026-10-28 |
|
| Workflow Orchestration | Full |
|
Review is split across specialist agents for logic and bugs, standards, architecture and risk that run in parallel with dedicated context, and a judge agent consolidates their findings, removes duplicates, resolves conflicts and filters by confidence; a finding recommendation agent then ranks each issue against PR history. Review effort mode, in research preview, routes each change to a review depth matched to its risk, and Spec, Design review and Remediation agents, all in research preview, check implementations against linked specs and Figma designs and open a separate remediation pull request for findings above a severity threshold. PR Triage, in research preview since Qodo 3.0, queues open pull requests across repositories and providers under SLA, priority and blast radius policies. Custom agentic quality workflows are an Enterprise feature. Sourcedocs.qodo.ai/core-concepts/qodo-platform-core-capabilities, docs.qodo.ai/whats-new and qodo.ai/pricingread 2026-10-28 |
|
| Knowledge Grounding & RAG | Full |
|
The Context Engine is the platform's knowledge layer and indexes and reasons over the customer's codebases, detecting cross module issues and dependencies across services and across Git providers. PR history indexing indexes each connected repository's pull request history in the background, keeps it current incrementally as pull requests merge and shows its status in the portal, so every finding is classified against how similar issues were handled before. Software Map, in beta, discovers repositories and keeps a live map of their dependencies, contracts and the blast radius of changes, updated with every pull request, and web search for dependency context, in research preview, pulls current documentation for third party libraries. Sourcedocs.qodo.ai/core-concepts/context-engine, /core-concepts/pr-history and docs.qodo.ai/whats-newread 2026-10-28 |
|
| Human Oversight & Guardrails | Partial |
|
Output is advisory, since reviews land as prioritized findings and suggested changes on a pull request that a human accepts or rejects, and Qodo says sitting outside the generation loop lets it check code independently rather than approve its own work. The Remediation agent, in research preview, opens its fixes as a separate remediation pull request for review rather than committing to the branch, mined rules can be held in a pending tab until an admin accepts them, and Review Standards and REVIEW.md let teams encode the rules agents apply, with governance analytics on Enterprise. Qodo has no per action approval gate or runtime guardrail of its own, since the agent does not act autonomously on the codebase, and it does not approve or merge pull requests. Sourceqodo.ai, docs.qodo.ai/whats-new and qodo.ai/solutions/enterpriseread 2026-10-28 |
|
| Security, Identity & Governance | Full |
|
Qodo's homepage says it is SOC 2 Type II certified through independent audit, with the trust center at trust.qodo.ai holding the details, and that code is analyzed and discarded under zero data retention with nothing used to train models. Access controls include SSO and SAML with both service provider and identity provider initiated sign in, roles and permissions, invite only access, restricting reviews to registered users, API key management and scoped context access, with audit logs on Enterprise. Sourceqodo.ai, docs.qodo.ai/whats-new (Qodo 2.6), docs.qodo.ai/account-management and qodo.ai/pricingread 2026-10-28 |
|
| Observability & Auditability | Partial |
|
Each finding lands on the pull request with an explanation, the relevant snippet, quality labels, evidence and reasoning and suggested remediation, with the evidence section linking to the rule or repository relationship behind it. A Findings page collects findings across the organization, the code review analytics dashboard, in beta, shows how teams respond to findings by repository and author with CSV export, and Enterprise adds audit logs and governance analytics; Qodo describes an audit trail of every issue and compliance flag. None of these is a per step or per tool call record of what the review agents did, since the findings are the output and Qodo does not say what the audit log records. Sourcedocs.qodo.ai/whats-new (Qodo 2.3 and 3.0), qodo.ai and qodo.ai/pricing; docs.qodo.ai/core-concepts/qodo-platform-core-capabilitiesread 2026-10-28 |
|
| Memory & State Persistence | Full |
|
Rule Miner mines the organization's pull request history for recurring accepted review patterns and writes rules from them, up to ten per repository on the first run and more every two weeks, scoped to the organization, repository or the paths they govern and traced back to the pull request each came from. Generated rules either activate directly or wait in a Pending rules tab for an admin to accept; rules can be edited, activated or removed afterwards; and once a rule is active, dismissed suggestions it generates reduce its signal over time so noise fades on its own. The rules persist because the system learned them, not because a person wrote them. Sourcedocs.qodo.ai/governance/rule-enforcement/rule-miner and docs.qodo.ai/whats-newread 2026-10-28 |
|
| Deployment & Data Residency | Full |
|
Deployment spans multi tenant SaaS, single tenant SaaS, on premises and fully air gapped environments so code never leaves the customer's infrastructure, with zero data retention where code is analyzed and discarded and no training on customer code; Gerrit and Bitbucket Data Center connect to on premises installs through guided wizards, and PR-Agent can be self hosted with the team's own model keys. Sourceqodo.ai, qodo.ai/pricing and docs.qodo.ai/whats-new (Qodo 2.9 and 3.0)read 2026-10-28 |
|
| Prebuilt Agents, Templates & Packs | Full |
|
Ready made capability comes in several named sets. The specialist review agents (logic and bugs, standards, architecture, risk, a judge agent and a finding recommendation agent) run out of the box, with Spec, Design review and Remediation agents in research preview; the Agentic Toolbox packages skills (Codebase Wisdom, Get Rules, Review, Review Resolver, Manage Standards) that install into Claude Code, Codex and Kiro as verified marketplace plugins or connect over MCP; and Review presets (Minimal, Standard, Comprehensive, or one built by the onboarding wizard) are ready made code review configurations a team picks and switches. Sourcedocs.qodo.ai/agentic-toolbox/agentic-toolbox-overview, /core-concepts/qodo-platform-core-capabilities and docs.qodo.ai/whats-newread 2026-10-28 |
|
| Triggers & Channel Coverage | Full |
|
Reviews fire automatically on pull request events across GitHub, GitLab, Bitbucket and Azure DevOps and on Gerrit change events for Enterprise, with the open source PR-Agent extending coverage to CodeCommit and Gitea. People reach Qodo by addressing it as qodo in a pull request comment or thread, through slash commands, from the Agentic Toolbox's review skill inside a coding agent before a pull request exists, and from the CLI in pre commit hooks and CI gates; the IDE plugins that ran pre commit audits retire on 30 October 2026. Sourceqodo.ai, docs.qodo.ai/whats-new and docs.pr-agent.airead 2026-10-28 |
|
| Model Flexibility & Routing | Full |
|
Bring your own key is an Enterprise feature, and Qodo's homepage names the providers it covers: the customer brings its own OpenAI, Anthropic, Azure OpenAI or self hosted models, so the customer chooses who runs inference. Review effort mode, in research preview, lets Qodo match review depth to the risk of each change or a repository hold a fixed mode. Sourceqodo.ai, qodo.ai/pricing and docs.qodo.ai/whats-newread 2026-10-28 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Other systems call Qodo through the Agentic Toolbox, generally available since Qodo 2.9, which connects coding agents, CI/CD pipelines, internal tools and custom applications over MCP or APIs, installs from a CLI, and ships as verified plugins on the Claude Code, Codex and Kiro marketplaces. API keys are managed in the account settings, and repository behavior is configurable through a .pr_agent.toml file, REVIEW.md and the portal. PR-Agent remains open source under community governance, separate from the commercial platform. Sourcedocs.qodo.ai/agentic-toolbox/mcp, /agentic-toolbox/cli, docs.qodo.ai/whats-new and /account-management/manage-api-keysread 2026-10-28 |
|
| Testing, Debugging & Optimization | Full |
|
Qodo is an AI code review and governance platform that runs specialist review agents on every pull request, and before it through the Agentic Toolbox's review skill inside a coding agent, and classifies each finding against the repository's history, so reviewing the customer's change is the product itself; blast radius labels, in research preview, rate each change Large, Medium or Small. Qodo publishes its own AI code review benchmark. Its current pages do not mention test generation or Qodo Cover. Sourcedocs.qodo.ai/core-concepts/qodo-platform-core-capabilities, docs.qodo.ai/whats-new and qodo.ai/pricingread 2026-10-28 |
|
| Browser & Computer Use | Not documented |
|
The platform works on repositories, pull requests and the IDE through git platform APIs and plugins, and the Design review agent compares code against linked Figma designs rather than driving a browser. Qodo offers no browser control or way to operate software that lacks a programmatic interface. Sourceqodo.ai and docs.qodo.ai/whats-newread 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
Repository access filtering in Qodo 3.2 lets enterprises enforce repository level access using their GitHub Enterprise Server or GitLab permissions, so each user only reaches the repositories they are authorized to see. The release also adds fix loop optimization, on by default: when a developer pushes fixes for Qodo's findings, the next review focuses on the recent changes instead of the whole pull request.
Bears on: Security / enterprise
View sourceOctober 30, 2026 is now the retirement date for Qodo's IDE plugin, formerly Qodo Gen, and the plugin points users to the Qodo Agentic Toolbox. From that date the chat panel and commit message generation show a retirement message instead of working, while on premises installations are not affected.
Bears on: Integrations
View sourceQodo 3.0 added PR Triage in research preview, one queue that pulls open pull requests from GitHub, GitLab, Bitbucket and Azure DevOps and orders them by policies for deadlines, priority, blast radius and inactivity. Each pull request shows its next action and how it relates to others across repositories.
Bears on: Workflow orchestration
View sourcePricing
Pro Team at $0.012 a credit, pooled across the team, in monthly packs of 2,500 credits ($30, about 18 reviews a month), 5,000 or 20,000, with no user limit and no annual commitment. A 14 day free trial comes first, and Enterprise, for 30 or more users, is custom.
Credits pooled across the team at $0.012 a credit, bought in monthly packs on Pro Team, with no seat price. Enterprise is a custom contract.
Included quota
Pro Team packs carry 2,500, 5,000 or 20,000 credits a month, pooled across the team, with no limit on repositories, users or reviews and no rate limits; developers do not need Qodo accounts. Credits expire at the end of each monthly cycle. Enterprise, pitched at 30 or more users, adds SSO and SAML, audit logs, governance and advanced analytics, advanced self learning, custom agentic workflows, Gerrit support, BYOK, single tenant SaaS or on premises and air gapped deployment, an MSA and DPA, and priority support with an SLA and a dedicated CSM.
What is public
Qodo offers a 14 day free trial with unlimited reviews and credits and no card. Pro Team is priced at $0.012 a credit pooled across the team. Packs hold 2,500 credits (about 18 reviews a month), 5,000 (about 36) or 20,000 (about 144). The plan is designed for teams of up to 30, with unlimited repositories, no rate limits and no annual commitment. A custom Enterprise tier for larger teams adds SSO and SAML, audit logs, governance analytics, advanced self learning, BYOK, single tenant SaaS or on premises deployment, and priority support with a dedicated CSM. The free Developer tier and the per seat Teams price are no longer offered, and there is no permanent free tier. Open source projects can use Qodo free through the Qodo for Open Source program.
Billing mechanics
Reviews and IDE or CLI use draw on a credit pool shared across the team at $0.012 a credit, bought in monthly packs that can be upgraded, downgraded or switched from the dashboard at any time. Credits expire at the end of each monthly cycle. When the trial ends, reviews pause until a plan is picked. Enterprise is contracted.
Cost watchouts
Credits are pooled and review size drives consumption, so the reviews per pack figures are estimates. Overage runs at the same rate up to a spending cap the customer sets. Enterprise is pitched at 30 or more users, and Gerrit, audit logs, BYOK and self hosting are Enterprise only.
Variable cost rationale
Qodo meters credits. Reviews and IDE or CLI use draw on one credit pool shared across the team at $0.012 a credit, bought as a monthly pack of 2,500, 5,000 or 20,000 credits. Larger changes take more credits per review, and model choice matters, since an Opus request uses five times the credits of a standard one. A busy month uses up the pack and moves the team into overage. Overage is charged at the same per credit rate and accrues up to a monthly spending cap the customer sets and can change at any time. Credits expire at the end of each cycle with no rollover, Pro Team has no annual commitment, and packs can be switched from the dashboard. PR-Agent, the open source version, can be self hosted with Docker and the customer's own LLM keys, which removes the per credit cost for teams with the DevOps capacity to run it.
Additional watchouts
Cost scales with review volume, not seats. A 2,500 credit pack covers about 18 reviews a month for the whole team, so busy repositories need the larger packs. Audit logs, BYOK and single tenant or on premises deployment are Enterprise only.
Overage / add-ons
Reviews keep running once base credits are used up, and overage is charged at the same per credit rate as the plan, with no premium and no penalty. Overage accrues against a monthly spending cap the customer sets and can change at any time, with a notice before the cap is reached. Credits reset each month with no rollover, and PR review pools are set per plan.
Sales call required
Mixed (some tiers require a call)
Free / trial
14 day free trial with unlimited reviews and credits, no card required. Free for qualifying open source projects.
Lowest paid plan
Pro Team at $0.012 a credit, from a 2,500 credit pack a month (about $30)
Commercial notes
Qodo was formerly CodiumAI and sells an AI code review and governance platform. Its pricing moved from per seat with a free Developer tier to a trial plus pooled credits.
Key ambiguities
Credits per review vary with the size of the change, so the cost per review (about 18 reviews a month on a 2,500 credit pack) is an estimate. Enterprise pricing is custom.
Cancellation / refund
Pro Team has no annual commitment, and credit packs can be upgraded, downgraded or switched from the dashboard at any time. Enterprise is a custom annual contract.
Support SLA / resale
Enterprise adds priority support with an SLA and a dedicated CSM, plus BYOK and single tenant SaaS or on premises deployment. Qodo does not train models on customer code.
Missing data
Credits per review vary with the change, so the per review cost is an estimate. Enterprise pricing is custom.
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 Qodo
The closest documented capability profiles to Qodo among coding agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Potpie AI11.0 / 14A lighter documented profile than Qodo
- Sonar11.0 / 14A lighter documented profile than QodoQodo vs Sonar →
- cubic10.5 / 14A lighter documented profile than Qodo
- Embedder11.5 / 14Fuller documented coverage on Human Oversight & Guardrails
- Augment Code12.0 / 14Fuller documented coverage on Human Oversight & Guardrails and Observability & Auditability
- CodeRabbit12.0 / 14Fuller documented coverage on Human Oversight & Guardrails and Observability & AuditabilityQodo vs CodeRabbit →
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded