The intent gets scattered.
The real task lives across a chat, a terminal, and the decisions you made along the way.
AN AI-NATIVE IDE · IN DEVELOPMENT
CODOVYN is an AI-native development environment being built around task continuity, flexible agent choice, and control over your workflow.
One IDE. Every Agent. Endless Flow.
import { db } from "../db";const PAGE_SIZE = 20;export async function getActivity(cursor?: string) { const rows = await db.activity.findMany({ take: PAGE_SIZE + 1, orderBy: { id: "desc" }, ...(cursor ? { cursor: { id: cursor }, skip: 1 } : {}), }); const items = rows.slice(0, PAGE_SIZE); return { items, nextCursor: rows.length > PAGE_SIZE ? items.at(-1)?.id ?? null : null, };}A stopped agent shouldn’t mean reconstructing the goal, retracing file changes, and explaining the same decisions all over again.
The real task lives across a chat, a terminal, and the decisions you made along the way.
A conversation alone doesn’t tell the next agent what changed or what still needs checking.
Before work continues, you need to know what happened—and what is allowed to happen next.
THE AGENT NETWORK / CONCEPT DEMONSTRATION
One agent works. The task has a center of its own.
A changing agent. An unchanged task.
One agent works. The task has a center of its own.
Availability can change. Intent, files, and decisions shouldn't disappear.
A new path carries the preserved context toward the next agent.
A possible continuation, with a review boundary and your permission.
Follow an illustrative task through an interruption. See the context a safer continuation could carry forward.
An agent is adding cursor pagination. The goal, changed files, and next check belong to the task.
import { db } from "../db";const PAGE_SIZE = 20;export async function getActivity(cursor?: string) { const rows = await db.activity.findMany({ take: PAGE_SIZE + 1, orderBy: { id: "desc" }, ...(cursor ? { cursor: { id: cursor }, skip: 1 } : {}), }); const items = rows.slice(0, PAGE_SIZE); return { items, nextCursor: rows.length > PAGE_SIZE ? items.at(-1)?.id ?? null : null, };}A pagination task is in progress. Intent and file changes stay attached.
The agent stops. Execution pauses before the next command.
Review the goal, file state, decisions, and remaining checks.
Choose an illustrative agent. Approve the next step yourself.
An interactive product vision, not a live agent session. Multi-provider continuation has not been verified here.
Keep the project, the conversation, and the next decision in view. Explore the editor, task snapshot, and proposed review state below.
2 illustrative file changes · nothing runs automatically
− take: PAGE_SIZE,+ take: PAGE_SIZE + 1,+ cursor: cursor ? { id: cursor } : undefined,+ skip: cursor ? 1 : 0, // Return a cursor only when another page exists.Stable cursor order reviewed
Empty-page behavior considered
Execution requires your approval
A familiar editor and project tree. No need to make your code the supporting act.
See the goal, decisions, and handoff boundary in the Task context tab.
Review proposed changes before an agent is permitted to continue.
This is an original interface concept, not a screenshot of released software. The controls explore a local simulation; no provider is connected.
CV / 004 — THE CONTEXT CORE
Agents come and go. The task needs somewhere to stay.
The goal and the acceptance checks.
The project at the continuation boundary.
The files this task actually depends on.
What was chosen, and the reason behind it.
The steps taken and what happened.
The changes you can inspect before continuing.
A conceptual model of preserved task context.
Provider handoff remains a product direction.
The vision is a workspace that keeps your project at the center—and leaves the decisions with you.
A flexible approach to agents and models, rather than a workflow tied to a single conversation.
A local-first development direction, with your files and editing workflow at the center.
Task intent, file state, decisions, and the next check—not just a longer chat history.
Clear permission boundaries and a chance to review before the next execution step.
CODOVYN is being built. This website shows the direction, not a finished IDE or a verified multi-provider session.
Follow the developmentExplore an illustrative workflow, inspect the proposed workspace, and join the email waitlist when the service is configured. The demonstration never contacts an AI provider or executes code.
Less starting over. More moving forward. Help shape a better home for AI-native development. Join for product updates and opportunities to explore the workflows as they take shape.
Still in development. Joining does not guarantee access or a launch date.
One email. A new thread to follow.