A new open-source tool released on September 14, 2026, is breaking down barriers between two of the AI coding world's biggest platforms. The cognition-claude-proxy allows developers to access Cognition's Devin model catalog directly through Anthropic's Claude Code interface, creating a unified workflow that lets teams experiment with multiple AI coding assistants without abandoning their existing tools.

What Problem Does This Proxy Actually Solve?

For developers evaluating different AI coding platforms, the friction has always been real. Switching between tools means learning new interfaces, reconfiguring keyboard shortcuts, and disrupting established workflows. The cognition-claude-proxy eliminates that friction by functioning as an API translation layer, intercepting Claude Code requests and routing them to Devin's model endpoints. This architecture lets developers already comfortable with Claude Code's interface access Devin's capabilities without switching tools or learning a new development environment.

The proxy handles the technical complexity behind the scenes, managing authentication, request formatting, and response translation between the two platforms. Developers can switch between Claude and Devin models within the same session, comparing outputs or selecting the best model for specific code generation scenarios. This flexibility is particularly valuable for organizations evaluating multiple AI coding assistants or those with specific use cases where Devin models demonstrate performance advantages.

How to Use the Cognition-Claude-Proxy in Your Workflow

• Install and Configure: The proxy is built in Python and requires API credentials from both Anthropic and Cognition to function. The GitHub repository includes installation scripts and example configurations for common development scenarios, making setup straightforward for most developers.
• Route Requests Intelligently: Real-time request routing with minimal latency overhead allows you to send Claude Code requests to Devin endpoints seamlessly. The proxy supports streaming responses from Devin models, maintaining the interactive experience developers expect.
• Track Usage Across Platforms: Automatic token usage tracking across both Anthropic and Cognition platforms helps teams monitor costs and understand which models are being used most frequently in their workflows.
• Select Models by Task: Configurable model selection through environment variables lets you choose different models for different coding tasks without changing your development environment or rewriting integration code.

The project is licensed under MIT, making it freely available for both personal and commercial use. The initial version (0.1.0) focuses on core proxy functionality, with the developer indicating plans for additional features like caching, rate limiting, and multi-user support in future releases.

Why This Matters for the Broader AI Coding Landscape

The cognition-claude-proxy represents a growing trend of community-built interoperability tools in the AI development space. As model providers proliferate and specialize, developers increasingly need middleware solutions that allow them to combine the strengths of multiple platforms. This proxy gives Claude Code users immediate access to Cognition's catalog without requiring platform migration, lowering the barrier to multi-model experimentation.

The tool also addresses vendor lock-in concerns by creating an abstraction layer between the development interface and the underlying model provider. Teams can experiment with different models without rewriting integration code or retraining developers on new tools. For organizations managing AI coding workflows, tools like this offer a pragmatic path to evaluating and integrating new capabilities while preserving existing investments in developer tooling and training.

The proxy has already received attention on Hacker News, where the developer posted a Show HN thread to gather community feedback. Early adopters are testing the proxy in production workflows, particularly for projects that benefit from comparing model outputs across providers. This real-world validation suggests the tool addresses a genuine pain point in how development teams currently evaluate and adopt AI coding assistants.

Meanwhile, Cognition itself continues to expand its offerings. The company recently launched Fusion, a new harness within Devin CLI that combines separate models for planning and execution, cutting costs by 39% across coding benchmarks compared to using a single frontier model. This cost-efficiency focus, paired with community tools like the cognition-claude-proxy, suggests Cognition is positioning Devin as both a capable and economical choice in an increasingly crowded market of AI coding assistants.