May 2026 Summaries

Zed's Agent Panel has introduced Terminal Threads, allowing users to manage terminal-based workflows as threads within the sidebar, enhancing the integration of tools like claude, amp, and codex into Zed's environment. This development follows the introduction of Parallel Agents, which supported Zed's built-in and ACP-connected external agents but did not accommodate terminal workflows efficiently. Terminal Threads enable users to open and manage as many terminal sessions as they need, track their progress, and receive notifications, thus offering a seamless experience for those who prefer using terminal agents. While ACP continues to provide deeper integration with Zed's editing workflows, Terminal Threads offer a flexible alternative, allowing users to maintain existing workflows without additional costs, especially with changes in Claude Code's subscription model. The introduction of Terminal Threads does not diminish Zed's commitment to ACP, as both features are intended to complement each other, providing users with versatile options for integrating external agents into their workspace.

Local models offer several advantages over cloud-hosted models, such as enhanced privacy, cost-effectiveness, and control over the system, making them attractive to developers who prefer to keep their data on personal hardware and avoid dependencies on SaaS platforms. Despite the benefits, local models are generally less powerful than cutting-edge cloud-based solutions and may require significant hardware resources to run efficiently. Tools like LM Studio, Ollama, and llama.cpp facilitate the use of local models, and Zed supports these tools, allowing developers to experiment with different configurations and models, such as the Qwen 3.6 35B A3B model, which employs a "Mixture of Experts" approach for faster performance. While local models demand more discipline and attention to be used effectively, especially due to smaller context windows and lower "cleverness," they offer a reliable and customizable alternative for those wary of changing cloud provider conditions.

Zed now allows users to sign in with their ChatGPT accounts to access OpenAI's models directly within the Zed platform, offering the same usage benefits as Codex, which is included with every ChatGPT subscription. This integration comes at a time when other platforms like Anthropic and GitHub Copilot are shifting towards usage-based billing, but Zed and OpenAI aim to provide cost-effective options through existing subscriptions. Users can configure the integration by selecting models in the agent panel and benefit from flexible usage options, including using Zed's built-in agent, an ACP adapter for clients like JetBrains, and running Codex via the CLI. OpenAI's support for subscription-based access contrasts with the industry trend towards usage-based billing, and Zed encourages users to explore their editor on various operating systems.

Anthropic has announced changes to the billing structure for Claude subscriptions, distinguishing between usage through its first-party tools and third-party agents like those used in Zed. Starting June 15, usage of Claude Code through third-party tools will no longer count toward existing subscription limits but will instead consume a new monthly "Agent SDK credit" allocated to each plan, with overages billed at standard API rates. This adjustment implies a significant cost increase for users heavily relying on agents, who previously benefited from subsidized rates. Zed, however, offers flexibility by supporting multiple providers and agent models, allowing users to maintain their existing Claude subscription limits by running Anthropic's official CLI in a terminal within Zed or opting for alternative agents. The open ACP protocol ensures users are not restricted to a single provider's pricing, enabling seamless transitions as billing models evolve. Zed remains a versatile platform for agentic development, offering various configurations and resources for users to adapt to these changes.

Zeta2.1, the latest update to Zed's edit prediction model, offers significant improvements in performance and efficiency compared to its predecessor, Zeta2. The new model emits significantly fewer output tokens, which enhances prediction speed by up to 50 milliseconds and reduces server usage by 30%. These advancements are attributed to a new prompt format called "Multi-Region," which focuses changes only around the code intended for edits, resulting in faster predictions with each keystroke. Zeta2.1 remains open-weight and is available for download on Hugging Face, having been trained on opt-in data from open-source repositories. The update also includes the publication of Rust code bindings to PyPI for easier self-hosting, making Zeta2.1 the default choice for Zed users, with options to explore Zed Pro or Zed Business for more extensive features. The model is designed to be run locally and is accessible on multiple operating systems, while Zed continues to seek passionate individuals to join their team in advancing software development.

Zed for Business, now available at $30 per seat per month, is designed to facilitate the integration of Zed within organizations by providing enhanced privacy controls and administrative oversight. It ensures that privacy settings like prompt sharing and Edit Prediction training are disabled by default at the organizational level, preventing individual engineers from altering these settings and thereby addressing security team concerns. Administrators can also disable unapproved features and set global token spend limits for teams using Zed-hosted models. Unlike competitors who mandate their hosted AI, Zed allows companies to use their existing AI contracts or preferred vendors, offering billing at provider cost plus 10% if Zed-hosted AI is chosen. Although this release doesn't include SSO or SOC 2 compliance, which some companies require, Zed is actively working on these features. Organizations interested in setting up Zed for Business or requiring larger contracts can contact the company for more details.

Zed recently introduced Parallel Agents, which has received positive feedback, though some concerns were raised about the company's focus on AI integration and the perceived worthlessness of AI code. Zed initially sold tokens at a loss to AI users but shifted to token-based billing to cover infrastructure costs. This change aligns with industry trends of subsidizing AI for future gains, despite Zed not being a large corporation. The company plans to release "Zed for Business" for easier team collaboration with straightforward pricing and centralized features. Despite skepticism about AI in IDEs, the company believes AI can enhance productivity by serving as a companion in coding, although human oversight remains crucial. Zed's vision includes a collaborative workspace where humans and AI work together, embodied by their DeltaDB synchronization engine designed for shared, real-time, and asynchronous codebase views. Although AI integration is necessary due to growing adoption, Zed remains committed to fostering collaborative editing, believing that teamwork leads to the most effective software development.