Codium AI’s Codiumate Agent To Complete Your Code With You

Codiumate, developed by Codium AI, introduces an innovative approach to software development by seamlessly integrating AI automation with human oversight to enhance productivity.

Highlights:

• Codium released Codiumate Agent, a semi-autonomous AI agent aimed at transforming the software development lifecycle.
• It performs various tasks including writing code, designing a detailed plan, creating documentation, identifying code duplicates, and performing in-depth testing.
• Developers can access it by installing the extension on Visual Studio Code to simplify development workflows and enhance productivity.

Meet Codiumate Agent

Codium AI has launched Codiumate-Agent to enable enterprises to reap the benefits of autonomous AI coding while still retaining human oversight and compliance with regulations and existing software. It enhances efficiency by automating coding tasks, while developers retain full control over the direction and quality of their work.

Most of the developers use AI tools to generate code. However, a developer’s role involves many tasks other than just writing code that are often overlooked. These include:

1. Creating documentation
2. Improving code
3. In-depth testing
4. Identifying code duplicates

While AI tools are helping developers write code quicker, concerns are being raised that this could make software quality worse. A study by The Consortium for Information & Software Quality in 2022 stated that the cost of poor software quality in the US has grown to $2.41 trillion. This is what Codium AI is trying to improve using its Codiumate Agent.

CodiumAI CEO and co-founder Itamar Friedman said the following about their vision for the agent:

“While coding assistants like GitHub Copilot are extremely popular, more and more developers are insisting on capabilities beyond what is available. Developers want AI to help them with the often tedious tasks that are critical to ensuring code quality and correctness, and these go beyond just writing code. We at CodiumAI are looking to an Agentic future where busy developers will team up with agents to improve coding quality and processes. Integrated agents will be there every step of the way, offering insights, recommendations and solutions proactively.”

-Itamar Friedman, CEO, Codium AI

Developers can access Codiumate Agent by installing its extension on Visual Studio Code. After installing this, users can use it for various tasks to simplify development workflows and enhance productivity.

What can Codiumate Agent do?

The Codiumate Agent provides an integrated and proactive solution to the challenges faced by developers about software quality. It takes existing code snippets that developers highlight in their environment and automatically designs a step-by-step development plan.

It writes code as needed, creates detailed documentation, suggests different tests, and identifies duplicates within the code to make sure that the code works properly before deploying it in a production environment.

With the help of these steps, there is the guarantee of higher code quality which will be beneficial to any software development team.

How Does It Work?

Friedman demonstrated the Codiumate tool live, showcasing how users can drag their mouse cursor to highlight multiple code blocks within their development environment. This allows them to instruct Codiumate to generate a plan for utilizing that code.

Codiumate Agent then presents the plan to the human user for review and allows the user to edit and adjust any portion of it as per their preferences. Once satisfied with the plan, the developer can click a button labelled “Start implementing the plan.”

This prompts the agent to begin with the coding phase. Throughout this coding phase, the agent occasionally takes pauses to confirm its actions with the developer. This allows it time to make adjustments in the code as needed.

It can also prompt the human developer to complete or write additional code to ensure that the development journey continues smoothly.

The steps involved are as follows:

1. Context collection: Select and add all the context that you think could be helpful for Codiumate Agent to understand and plan the task.
2. Planning prompt: Describe your plan properly. Always give a detailed description of the plan. Sometimes, if you don’t add enough description, Codiumate-Agent might hint that you should add more relevant info.
3. Refine the plan (Optional)
4. Generating a detailed plan (optional)
5. Plan-aware auto-complete for implementation: The developer starts opening files one by one according to the suggested plan and uses Codiumate auto-complete. But this time, the developer will get an auto-complete++ since the agent knows what you are trying to achieve without too much of a hassle.
6. Receive suggestions on code smell, best practices, and issues
7. Use /quick-test, /commit, /describe

Underlying Technology

The Codiumate agent’s backend technology leverages top-tier OpenAI models. Alongside this, experiment trials are also being conducted for Anthropic’s Claude and Google’s Gemini. Codium also provides its own Large Language Model (LLM) through the AlphaCodium process, which boosts the capabilities of other LLMs in code completion tasks.

While OpenAI’s models are accessible to all users, Codium’s exclusive LLM is reserved for enterprise users who pay and subscribe. Friedman highlighted its effectiveness in coding tasks and noted that a Fortune 10 company, whose identity remains confidential, is already utilizing it in their production environment.

Codium practices zero data retention so that users are assured of data privacy. To address a growing market of AI tools that seek to offer security and anonymity to customers,

Codiumate also practices zero data retention, meaning that any code or information entered into Codiumate will immediately be deleted after the developer ends the session or closes the tool.

Conclusion

Codium AI’s Codiumate Agent is the first step, along with Devin AI, towards tandem development by automating coding tasks while empowering developers to retain control over the codebase.