Launched this week
Capacitor
Shared memory for coding agents
10 followers
Shared memory for coding agents
10 followers
Agents code fast. Capacitor helps your team understand them just as fast. It's a shared session layer for coding agents: hand off sessions from Claude Code to Codex, learn from past mistakes with an eval loop, review PRs with the agent's reasoning, not just the diff through a dashboard, CLI, and MCP tools your agents query directly.
Free
Launch Team / Built With
Hi I am Lokhesh and I am part of the Product and AI team at Kurrent.io and I would like to introduce https://capacitor.kurrent.io/
Capacitor records every coding agent session automatically. No separate record commands. No configuration per session. Every turn, tool call, test run, and reasoning block is streamed to the team’s Capacitor server in real time, indexed for full-text search, and linked to the repository and pull request it belongs to. One simple setup, and your agents have memory.
Six core capabilities are available:
Collaboration - Developers share live session links via tools like Slack. A teammate opens the link, reviews the spec or reasoning in progress, and contributes directly without reconstructing context from message fragments.
Multi-agent handoffs - Capacitor enables real-time coordination across multiple agents working in parallel or in sequence handing off, handing back, and continuing work fluidly as the task evolves. Any agent can pick up exactly where another left off, with full context of what was done, what was decided, and what was already tried. Switch models mid-task, run agents in parallel on different workstreams, or bring in a second agent when the first hits a wall Capacitor keeps the entire team, human and agent, in sync.
PR review with session context - Review agents can pull the sessions behind a pull request at review time, giving reviewers and their agents access to the tests, attempts, and reasoning that produced the final diff. Changes become immediately explainable.
Evaluations and institutional learning. Capacitor scores sessions against team-defined rubrics correctness, test fidelity, surface area, time on task and promotes durable findings into per-repository guidelines. Those guidelines are automatically loaded at the start of the next agent session, so lessons from one session become guardrails and best practices for every session that follows.
Multi-player sessions - Capacitor turns solo agent sessions into team sessions. Launch any agent from the dashboard, share the link, and your whole team is in driving, contributing, and building together in real time from any browser.
Active session memory - Humans and agents can search the team’s full session history through built-in MCP tools. When a developer asks “have we worked on this before?” the agent searches, summarizes, and drills into the exact point where the relevant decision was made surfacing prior context in the current chat instead of reopening closed questions.
This is a private preview so you will need to join waiting list.
Thanks, let me know what you think!