Coder has launched Blink, an open source engine designed to help developers build, test, and deploy AI agents directly from the terminal.
Blink was developed in response to growing interest and demand for AI agents, which are used across industries to improve personalisation and operational efficiency. Despite the surge in interest, research indicates that a significant proportion of AI agent projects fail to reach production. According to Gartner, more than 40% of agentic AI projects may be discontinued by 2027, mostly due to excessive costs and challenges in execution.
Coder, which focuses on AI development infrastructure, is aiming to address these challenges with Blink. The offering includes an open source command-line interface (CLI) for developers to create agents and Blink Cloud, a hosted platform for deployment and collaboration within teams. For its Early Access release, Blink provides support for Slack integration, allowing users to develop bots that can interact within the Slack platform. Integration with GitHub and additional environments is planned for the near future.
Kyle Carberry, Co-founder at Coder, said, "It's surprisingly difficult to build and personalize even the most basic AI agents, such as a SlackBot or a GitHub code reviewer. Blink brings AI to the tools where chat already happens and makes development simple: no complex workflow builders, just agents as code."
The traditional development of AI agents can involve overcoming hurdles related to system complexity, integration of data, and acquiring the necessary skillsets. Many organisations either abandon or postpone their agent projects as a result. Blink has been designed to minimise these friction points by providing a workflow centred on the command line and direct programming, avoiding the need for visual workflow builders.
With Blink, developers can work entirely from the terminal, making the process more efficient and in line with established developer habits. They are able to test and refine AI agents locally, observing the behaviour in real-time, and then choose to deploy the agent to the cloud once satisfied with its operation. This approach is intended to reduce complexity and overhead, offering a code-first framework that supports any model, API, or deployment infrastructure.
In its current release, Blink offers several key components: the Blink CLI for building and testing agents from the terminal, Blink Cloud for team deployment and sharing, and Slack support for chatbot activity within that platform. Early Access is currently available at no cost, with the company encouraging contributions and feedback from the developer community to help steer future development.
"It's surprisingly difficult to build and personalize even the most basic AI agents, such as a SlackBot or a GitHub code reviewer," said Carberry. "Blink brings AI to the tools where chat already happens and makes development simple: no complex workflow builders, just agents as code."
Blink's release was informed by feedback from the developer community and by lessons learned through Coder's earlier open source and proprietary AI development solutions. The company noted that its existing enterprise development environments already offer features to support AI and tooling workloads, including GPU access, modular templates, and secure workspaces intended for agent and software development.
Blink is intended for developers specifically seeking practical AI agents without becoming tied to a particular platform or service. By being open source and code-first, it aims to allow for rapid iteration, transparency, and adaptability to a wide range of use cases and business requirements. Coder has positioned Blink to meet developers in the tools they already use and is planning additional integrations to expand the range of systems with which Blink can interact.
The Blink Early Access programme is free to download, and Coder has expressed an intention to collaborate with developers to define the evolution of the platform.
