CtrlK
Qodo (formerly CodiumAI or Codium) is an AI-powered code integrity platform that provides automated code review, test generation, and code quality tools for software developers. Founded in 2022 by Itamar Friedman and Dedy Kredo in Tel Aviv, Israel, the company originally focused on AI-driven test generation before expanding into a comprehensive platform covering the full software development lifecycle. In September 2024, CodiumAI rebranded to Qodo, a name combining "Quality" and "Code," alongside a $40 million Series A funding round that brought total funding to approximately $50 million [1][2].
Qodo's product suite includes Qodo Gen (an IDE plugin for code and test generation), Qodo Merge (an AI-powered pull request review system), and Qodo Command (a CLI-based agent framework). The company's open-source PR-Agent project, which automates pull request analysis, has accumulated over 10,700 stars on GitHub [3]. As of early 2025, Qodo's tools had been used by more than one million developers, and its enterprise platform had been adopted by multiple Fortune 100 companies [2].
Qodo is distinct from Codeium (now Windsurf), a separate company with a similar-sounding name that focuses on code completion and acceleration. The naming overlap caused frequent confusion among developers, which was one factor motivating the rebrand [4].
CodiumAI was founded in 2022 by Itamar Friedman and Dedy Kredo. Friedman, the company's CEO, had previously co-founded Visualead (a visual QR code startup acquired by Alibaba Group) and served as Director of Machine Vision at Alibaba's Israeli AI Lab. He holds a Master's degree in Computer Vision and a Bachelor's degree in Electrical Engineering from the Technion (Israel Institute of Technology). Kredo, who serves as Chief Product Officer, had previously led product and data science teams at Exploriem and VMware [5][6].
The company's founding thesis centered on the observation that while AI code generation tools like GitHub Copilot were making it faster to write code, the tools available for verifying and testing that code had not kept pace. Friedman and Kredo saw an opportunity to apply large language models to automated test generation, helping developers catch bugs and logic errors before code reached production [7].
CodiumAI launched its closed alpha in January 2023. Within weeks, thousands of developers had installed the CodiumAI IDE extension. The product generated behavior-specific unit tests by analyzing source code and identifying edge cases, an approach that differentiated it from general-purpose code completion tools [7].
In March 2023, CodiumAI raised $11 million in seed funding co-led by TLV Partners and Vine Ventures. The round reflected investor interest in AI tools that focused on software quality rather than code generation alone. At the time of the seed round, the company was growing its user base rapidly and had begun developing its PR-Agent open-source project [8].
Throughout 2023 and into 2024, CodiumAI expanded from its original test-generation focus into a broader code integrity platform. The company released PR-Agent as an open-source tool for automated pull request review, which quickly gained traction in the developer community. The company also launched Codiumate, an IDE plugin that added code completion, explanation, and refactoring capabilities alongside the original test generation features [3].
By mid-2024, CodiumAI's VS Code extension had reached over 844,000 users, its JetBrains plugin had been installed more than 613,000 times, and its GitHub Marketplace listing had over 10,600 users [9].
On September 30, 2024, the company announced its rebrand from CodiumAI to Qodo alongside a $40 million Series A funding round. The round was led by Susa Ventures and Square Peg Capital, with participation from Firestreak Ventures, ICON Continuity Fund, and existing investors TLV Partners and Vine Ventures. The round was oversubscribed, bringing the company's total funding to approximately $50 million [1][2].
The rebrand served multiple purposes. The name "Qodo" was chosen to reflect the company's broadened mission beyond test generation, combining "Quality" and "Code" to signal its focus on code integrity across the full development lifecycle. The name change also resolved persistent confusion between CodiumAI and Codeium, a different company operating in the AI coding space. As part of the rebrand, Codiumate was renamed to Qodo Gen and PR-Agent's commercial offering became Qodo Merge [9].
At the time of the rebrand, CEO Itamar Friedman stated that the Codium name would continue to represent the core foundation models powering the platform's code integrity capabilities [9].
On March 11, 2025, Qodo released Qodo Gen 1.0, a major update that introduced agentic workflows to the IDE plugin. The release added an Agent Mode in which the AI could autonomously break down complex tasks into steps, select appropriate tools, and execute multi-step problem-solving without manual intervention. This release also introduced support for Anthropic's Model Context Protocol (MCP), enabling integration with enterprise tools such as Jira, Git, APIs, and databases [10].
In June 2025, Qodo launched Qodo Command, a CLI-based agent framework that allows developers to create, customize, and deploy AI coding agents from the terminal. Qodo Command supports multiple execution modes, including CI mode for automated validation and webhook mode for triggering agents via HTTP endpoints. The tool includes pre-built agents for common tasks such as code review, test coverage analysis, and release notes generation [11].
Qodo Gen (formerly Codiumate) is an IDE plugin available for Visual Studio Code and JetBrains IDEs (IntelliJ IDEA, PyCharm, WebStorm, and others). It provides AI-assisted code generation, test generation, code completion, debugging, and code explanation [10].
Qodo Gen offers two distinct operating modes [10]:
| Mode | Description | Best For |
|---|---|---|
| Standard Mode | Direct command-response execution without autonomous tool selection. Users manually add repos, files, and folders as context. | Simple queries, specific code generation, retrieving information |
| Agent Mode | Autonomous multi-step problem-solving. The AI analyzes requests, identifies needed tools, gathers context, and applies tools iteratively. | Complex tasks requiring reasoning across multiple data sources, full-stack code generation, bug investigation |
Test generation was Qodo's original core feature. Using the /test command in the IDE, Qodo Gen guides developers through generating a complete test suite, including example tests, mock configurations, and edge case behaviors. The system analyzes source code to identify potential failure modes and generates tests that target those specific behaviors rather than producing generic test templates [10].
As a developer types, Qodo Gen analyzes code in real time to infer intent and presents code completions as grayed-out suggestions directly in the editor. The completion engine draws on the indexed codebase for context, allowing it to suggest code that follows existing patterns and conventions in the project [10].
Qodo Gen 1.0 integrates with Anthropic's Model Context Protocol (MCP), allowing the AI agent to access external tools and services during agentic workflows. This enables scenarios such as retrieving Jira issue details, fetching Git commit history, and triggering CI/CD pipelines within a single session. Qodo Gen was the first MCP-enabled AI coding agent for JetBrains IDEs [10].
Qodo Merge is Qodo's commercial AI-powered pull request review product. It is the hosted, enterprise-grade version of the open-source PR-Agent project. Qodo Merge integrates with GitHub, GitLab, Bitbucket, and Azure DevOps, scanning pull requests with full organizational and historical context [3][12].
Qodo Merge provides automated code review through a set of tools that can be triggered via PR comments or configured to run automatically [3][12]:
| Tool | Command | Function |
|---|---|---|
| Describe | /describe | Generates automated PR descriptions summarizing changes |
| Review | /review | Provides AI-powered code analysis identifying bugs, security issues, and logic errors |
| Improve | /improve | Suggests code enhancements and optimizations |
| Ask | /ask | Answers natural language questions about the PR code, including line-specific queries |
| Add Docs | /add_docs | Generates documentation for code changes |
| Generate Labels | /generate_labels | Creates automated tags and labels for PRs |
| Similar Issues | /similar_issues | Identifies related issues in the repository |
| Update CHANGELOG | /update_changelog | Maintains change logs automatically |
| Compliance | /compliance | Runs checks for security, ticket requirements, codebase duplication, and custom rules |
Qodo Merge uses a PR Compression strategy that allows it to handle pull requests of any size. Each tool invocation uses a single LLM call (typically completing in approximately 30 seconds), keeping costs low. The system employs adaptive compression strategies, dynamic context retrieval, ticket context fetching, metadata management, and self-reflection mechanisms to optimize the quality of LLM outputs [3].
The commercial Qodo Merge offering includes features not available in the open-source PR-Agent, such as multi-repository awareness, organizational best practice learning, enhanced privacy controls, and custom compliance rule enforcement. Enterprise deployments can process over 20,000 PRs daily across customer organizations [12].
Qodo Command is a CLI-based agent framework launched in June 2025. It allows developers to create custom AI agents through configuration files that specify triggers, tool access, behavioral instructions, and output formats [11].
| Feature | Description |
|---|---|
| Custom Agent Creation | Define agents via configuration files specifying triggers, tools, instructions, and output formats |
| CI Mode | Automated validation with machine-readable output compatible with GitHub Actions, Jenkins, and other CI/CD systems |
| Webhook Mode | Turns agents into HTTP endpoints that external systems can trigger via POST requests |
| Web UI | Browser-based interface launched via qodo --ui, providing chat-based interactions, code diffs, and structured outputs |
| Multi-Agent Workflows | Sub-agents can invoke other agents as tools during execution, with activity surfaced in the parent agent's UI |
| Pre-built Agents | Includes agents for code review, test coverage analysis, and release notes generation |
Qodo Command is built on LangGraph for orchestrating agentic workflows and managing state across multi-step processes [11].
PR-Agent is Qodo's open-source pull request review tool, available under the AGPL-3.0 license on GitHub. It serves as the community edition of the technology that powers Qodo Merge. As of early 2026, the repository has over 10,700 stars, 1,400 forks, and 207 contributors [3].
PR-Agent supports five git hosting platforms: GitHub, GitLab, Bitbucket, Azure DevOps, and Gitea. It can be deployed via CLI, GitHub Actions, Docker, or self-hosted webhooks. The tool supports multiple AI models, including OpenAI GPT models, Claude, and DeepSeek [3].
Qodo Cover (formerly Cover-Agent) was an open-source tool for automated test generation and code coverage enhancement. It used large language models to analyze source code, generate test cases, validate that each test runs successfully, and verify that the new tests increase overall code coverage. Qodo Cover was inspired by Meta's TestGen-LLM paper and represented the first open-source implementation of that approach [13].
The open-source Qodo Cover repository is no longer actively maintained, but Qodo Cover Pro is available as a commercial product in preview, currently supporting Python projects [13].
Qodo frames its platform around the concept of "code integrity," which it defines as the ongoing ability of a codebase to remain correct, understandable, maintainable, testable, compliant, and safe as it evolves. Rather than focusing solely on code generation (as many AI coding tools do), Qodo positions itself as a platform that validates and reviews code at every stage of the development process [14].
Qodo's platform uses a multi-agent architecture in which specialized agents handle different aspects of code analysis. According to the company, no single model can hold enough specialized knowledge to handle all review tasks without tradeoffs, so Qodo deploys approximately 30 different middleware modules, each focused on specific aspects of code review and testing. Each agent is designed to know what to look for and what to ignore, which the company claims enables high recall (catching real issues) without generating excessive false positives [14].
Qodo developed a specialized code embedding model trained on NVIDIA DGX infrastructure. This embedding model powers the platform's retrieval-augmented generation (RAG) pipeline, enabling the AI to understand and analyze code semantics beyond simple text matching. The pipeline continuously maintains a fresh index of the codebase by retrieving files, chunking them into segments, and adding natural language descriptions to embeddings [15].
Qodo's platform is model-agnostic and supports multiple commercial LLMs. Supported models include OpenAI's GPT-4 and GPT-4o, Anthropic's Claude (including Opus and Sonnet variants), Google's Gemini, and DeepSeek models. Different models consume different numbers of credits in Qodo's billing system, with premium models like Claude Opus costing 5 credits per request [16].
Qodo uses mutation testing techniques to evaluate the quality of generated tests. Mutation testing works by introducing small changes (mutations) into the source code and checking whether the generated tests detect those changes. This approach helps ensure that the tests Qodo generates are meaningful and effective rather than superficial [14].
Qodo supports all major programming languages, with full feature support (test generation, code suggestions, code explanation, and behavior coverage) for the following languages [17]:
| Language | Test Generation | Code Suggestions | Code Explanation |
|---|---|---|---|
| Python | Full | Full | Full |
| JavaScript | Full | Full | Full |
| TypeScript | Full | Full | Full |
| Java | Full | Full | Full |
| C++ | Full | Full | Full |
| Go | Full | Full | Full |
| Ruby | Full | Full | Full |
| PHP | Full | Full | Full |
| C# | Full | Full | Full |
| Kotlin | Full | Full | Full |
| Swift | Full | Full | Full |
The platform also supports framework-specific patterns for popular ecosystems such as React, Django, and Spring Boot, as well as infrastructure-as-code formats including Terraform and Kubernetes YAML [17].
| IDE | Plugin Name | Users (as of Sept 2024) |
|---|---|---|
| Visual Studio Code | Qodo (formerly CodiumAI) | 844,600+ |
| JetBrains IDEs | Qodo (formerly Codiumate) | 613,000+ |
| Provider | Qodo Merge | PR-Agent (Open Source) |
|---|---|---|
| GitHub | Supported | Supported |
| GitLab | Supported | Supported |
| Bitbucket | Supported | Supported |
| Azure DevOps | Supported | Supported |
| Gitea | Not available | Supported |
Qodo offers a tiered pricing structure with free, team, and enterprise options [16]:
| Plan | Price | PR Reviews | IDE/CLI Credits | Key Features |
|---|---|---|---|---|
| Developer (Free) | $0/month | 30 PRs/month | 75 credits/month | PR code review, IDE plugin, CLI tool, community support |
| Teams | $30/user/month | Unlimited PRs (promotional) | 2,500 credits/user/month | All Developer features plus bug detection, PR automation, repo best practices learning, enhanced privacy, standard support |
| Enterprise | Custom pricing | Custom | Custom | All Teams features plus context engine, dashboard analytics, admin portal, MCP tools, SSO, priority support (2-business-day SLA), on-premise and air-gapped deployment options |
Credits reset every 30 days from the date the first message is sent, rather than on a calendar-month basis. Most operations consume 1 credit per LLM request, though premium models cost more (for example, Claude Opus costs 5 credits per request and Grok 4 costs 4 credits per request) [16].
Enterprise customers can deploy Qodo on-premises, in VPC environments, or in fully air-gapped setups with no internet connectivity. This makes Qodo suitable for regulated industries with strict data residency requirements. Enterprise deployments can also use proprietary self-hosted Qodo models [16].
| Round | Date | Amount | Lead Investors | Other Investors |
|---|---|---|---|---|
| Seed | March 2023 | $11 million | TLV Partners, Vine Ventures | Other undisclosed investors |
| Series A | September 2024 | $40 million | Susa Ventures, Square Peg Capital | Firestreak Ventures, ICON Continuity Fund, TLV Partners, Vine Ventures |
| Total | ~$50 million |
The Series A round was reported as oversubscribed. By 2025, the company had grown to approximately 100 employees distributed across Israel, the United States, and Europe [1][2].
Qodo's enterprise platform has been adopted by multiple Fortune 100 companies. Publicly reported customers include monday.com, Ford, Intel, Intuit, and MercadoLibre [2][18].
monday.com is one of Qodo's most publicly documented enterprise customers. According to case study data, monday.com uses Qodo to prevent over 800 potential issues per month while maintaining a 73.8% acceptance rate on AI-generated code suggestions [18].
A deployment with a Global Fortune 100 retailer reportedly saved over 450,000 developer hours in a single year, with individual developers saving approximately 50 hours per month. Across all enterprise customers, Qodo processes over 20,000 pull requests daily [18].
Qodo operates in the AI-assisted software development market alongside several categories of competitors [19]:
| Category | Tools | Primary Focus |
|---|---|---|
| Code completion and generation | GitHub Copilot, Tabnine, Amazon Q Developer, JetBrains AI | Inline code suggestions and autocomplete within IDEs |
| AI-native code editors | Cursor, Windsurf | Full IDE experience with integrated AI for multi-file editing |
| AI coding agents | Claude Code, Devin, Aider | Autonomous multi-step coding tasks and repository-level changes |
| Code review and quality | Qodo, Codacy, SonarQube | Code review automation, test generation, and quality enforcement |
Qodo differentiates itself from code generation tools by focusing on what happens before code is merged. While tools like GitHub Copilot and Cursor help developers write code faster, Qodo is designed to validate pull requests with context-aware analysis, enforce coding standards, and reduce code review risk at scale. Every AI-generated suggestion in Qodo includes linked commit history, relevant documentation, and matching test cases [19].
Unlike most competitors, Qodo provides an open-source foundation through PR-Agent, allowing developers and organizations to self-host and customize the PR review tooling. The commercial Qodo Merge product adds enterprise features such as multi-repository awareness and organizational best practice learning on top of this open-source base [3].
Qodo (formerly Codium/CodiumAI) is frequently confused with Codeium, a separate company that was later rebranded as Windsurf. Despite their nearly identical-sounding names, the two companies are entirely different entities with different products, teams, and technology. Qodo focuses on code integrity (testing, review, and quality), while Codeium/Windsurf focuses on code acceleration (completion and generation). The persistent confusion between the two companies was one of the factors that motivated CodiumAI's rebrand to Qodo in September 2024 [4].