Kurrent launches Capacitor private preview for shared memory across coding assistants

Kurrent announced the private preview launch of Kurrent Capacitor, a tool the company describes as shared memory for software teams working with AI coding assistants.

A shared history for human and AI development work

In a Business Wire announcement, Kurrent said Capacitor is intended to help developers and coding assistants work from the same record of prior activity. The company describes the product as an “AI-native shared memory” system for tools including Claude Code, Codex and Cursor.

The basic idea is to capture development sessions so that humans and coding assistants can later search, review and reuse the context behind code changes. According to Kurrent’s announcement, Capacitor provides searchable team session history and Model Context Protocol, or MCP, tools that can be used by coding assistants.

Kurrent’s GitHub repository for the kcap-cli command-line tool gives a more concrete view of the product’s scope. The repository says Capacitor can record Claude Code sessions, visualize coding-assistant activity in real time, support Codex and Cursor hooks, and provide pull request review tools grounded in development transcripts.

Why session memory matters

AI coding tools often operate inside a narrow context window. Developers may need to restate project conventions, architectural decisions or the reasoning behind earlier changes when moving between assistants, editors or review stages. Kurrent’s announcement positions Capacitor as a way to preserve that information as a team resource rather than leaving it scattered across individual chats or terminals.

The company says the product is currently in private preview, so availability is limited. The Business Wire release does not provide general-availability timing, pricing or customer adoption figures. It also does not include independent performance comparisons with other developer tooling.

Because the announcement comes from Kurrent, claims such as “first-of-its-kind” should be read as the company’s positioning rather than an independently verified market assessment. What is verifiable from the cited materials is that Kurrent has announced a private-preview product focused on recording, searching and exposing coding-session context to AI development tools.

Tooling details from the public repository

The public kurrent-io/kcap-cli GitHub repository describes a command-line interface for interacting with Capacitor. The repository says the CLI can record Claude Code sessions and includes hooks for Codex and Cursor. It also describes features for real-time visualization of assistant activity and pull request review based on recorded development transcripts.

Those repository details align with Kurrent’s broader announcement: Capacitor is framed less as another code-generation model and more as supporting infrastructure around coding assistants. Its role is to preserve activity history and make that information available to both developers and AI tools.

What remains to be seen

The private-preview status means there is still limited public evidence on how Capacitor performs in everyday engineering environments. Important open questions include how teams will manage privacy and retention for recorded development sessions, how well the system handles large codebases, and how developers will balance useful memory with sensitive information in transcripts.

For now, Kurrent’s launch adds to a growing category of developer tools built around AI-assisted programming workflows. Rather than focusing only on code generation, Capacitor targets the surrounding context: what was attempted, why decisions were made, and how that history can be reused during future implementation and review work.