The first IDE built for agents and teams.
Recursive gives agents the same workspace you use — sessions, tasks, an inbox, automations — so you can hand off work and pick up the baton when it matters.
Adapter and model agnostic. Start a session in Claude Code, pick it up in Codex, or let an automation spawn one. Every session keeps full context — plans, artifacts, event logs — no matter which model is driving.
Your backlog is the agent's backlog. Create tasks, prioritize them, break them into subtasks. Agents pick up work, update status, and report progress — all in the same task board you use.
A shared communication channel between you and your agents. When an agent needs a decision, it posts to your inbox with structured options. No more scrolling through chat logs to find what needs your attention.
Schedule recurring work or trigger agent runs from events. Daily code reviews, weekly dependency audits, post-merge cleanup — set it once, agents handle the rest.
Define how agents work, step by step. Research, plan, build, test, review — or create your own pipeline. Workflows are versioned, composable, and shared across your team.
Extend Recursive with MCP servers, custom skills, rules, and integrations. Connect to Slack, GitHub, your CI pipeline — agents use the same extensions you configure.
Common questions
What is Recursive?
Recursive is an autonomous development platform — think of it as the IDE layer between you and your AI agents. You give agents structured context (tasks, workflows, memory) and they do the work. You review the results and get things over the finish line.
How is this different from Cursor or Copilot?
Cursor and Copilot are excellent coding assistants that work alongside you. Recursive is the workspace around them. It manages what agents work on, how they work, and how their output gets back to you. You can use Cursor as your editor and Recursive as the operating system for your agents.
Does it just push code without my approval?
No. Every change happens in an isolated worktree on a separate branch. Agents post to your inbox when they need decisions. Nothing touches your main branch until you explicitly review and merge.
What kind of tasks can it handle?
Features, bug fixes, refactors, test coverage, documentation — anything you'd put in a ticket. Agents work best with well-scoped tasks, but can break down larger goals into subtasks automatically.
Is it locked to a specific AI model or editor?
No. Sessions are adapter and model agnostic. Start work in Cursor with Claude, continue in VS Code with GPT, or let an automation run with whatever model you configure. Recursive doesn't care — it manages the session, not the model.
How do I get access?
We're running a private beta. Drop your email below and we'll reach out when there's a spot for you.
Give your agents a proper workspace
Free for personal use. Lock in the intro price and get early access builds, remote access, and a commercial license.
Get Started