More agents. Less juggling.

Run Codex, Claude Code, Grok, OpenCode, Kimi Code, and Antigravity side by side. One desktop workspace for your projects, parallel tasks, and code review.

Building with agents? Join the Discord

A lot going on.
All in one place.

See how the code connects.

Explore your project and trace the reach of a change.

How it works

Browse the interactive codebase map, follow resolved file dependencies, and inspect related files during review. Use the map to choose what to check next; it does not replace tests.

Explore the codebase map

Give every agent room to work.

Run independent tasks side by side in separate Git worktrees.

How it works

Assign Codex, Claude Code, Grok, OpenCode, Kimi Code, or Antigravity to each task. Set dependencies when one change needs another, and follow progress without checking a pile of terminals.

Explore parallel work

Brief once. Build on it.

Carry project instructions and lessons into the next task.

How it works

Save project guidance once. New tasks match relevant guidelines to your prompt and inherit project defaults. Adjust the selection when needed and inspect what the agent received.

Explore project context

Your project. Your final say.

Review the changes and checks together before you merge.

How it works

Read each result beside its original brief. Combine related patches for review, ask for another pass, and decide what enters your project. Checks stay tied to the code they tested.

Explore code review

Work and personal, sorted.

Use the right agent account for every project.

How it works

Create named sign-in profiles and choose project defaults. Track reported task usage by project and account. Profiles organize sign-ins; your provider policies and local file permissions still apply.

Explore accounts and usage

Computer use & agent browser.

Let agents browse the web and use a Windows app you approve.

How it works

Built-in agent-browser lets agents navigate pages, fill forms, capture screenshots, and check accessibility in a separate browser session for each task. On Windows, choose an app window for clicks, typing, and scrolling. A visible status bar lets you pause or cancel control; moving your mouse or typing pauses it, and Escape cancels access. Add project tools through MCP connections.

Explore browser and computer use

Put repeat work on repeat.

Schedule checks and chores, with a result for every run.

How it works

Run on a schedule, or check a local branch and launch only when committed code changes. Change checks use no model calls. Review each run in its own history; Jackalope must be open and your computer awake.

Explore recurring tasks

Bring your
whole setup.

Multiple agents. Work and personal accounts.

Build a different lineup for every project. Choose allowed agents and default accounts, with managed profiles for Codex, Claude Code, Grok, and OpenCode.

Explore account profiles

Route the work. Keep the momentum.

Let your default agent choose an agent, model, and account from your project’s allowed lineup. Routing considers reported capacity; quota-aware handoffs preserve the work when an eligible fallback is available. Prefer a specific setup? Assign it yourself.

How routing works today

MCP, browser, and native desktop tools.

Discover MCP tools and connect them to your projects. Give agents built-in browser actions, screenshots, and local checks. On Windows, explicitly grant a desktop window for accessibility snapshots, clicks, typing, and scrolling.

See tool support and boundaries

Cross-agent communication, built in.

Give agents a shared view of active project work, ownership, scopes, and dependencies. They can send direct messages or project broadcasts; relevant updates arrive at launch, continuation, and Jackalope tool checkpoints.

How agents stay informed

One idea.
More ways forward.

Build the feature. Chase the bug. Explore another direction. Give independent tasks their own agents and worktrees, then decide what comes together.

“Add keyboard search and polish the settings.”

Project instructions + selected tools
Build keyboard searchCodex · Worktree 01
Polish the settingsClaude Code · Worktree 02
Search.tsx
+ onKeyDown={navigate}
+ onClose={returnFocus}
Combined review

Illustrated workflow

Meet your new workspace.

Track tasks across your projects.

Capture a thought. Start a task in a local project. See what is running, what needs an answer, and what is ready for review.

Silent demo with fictional Atlas project data. Switch between the board and list, then draft a brief for a new task.

Make room.
Make it yours.

Try a palette. The whole page comes along.
Pick your kind of light or dark.

Electric Indigo

Questions

What is Jackalope?

A desktop workspace for Codex, Claude Code, Grok, OpenCode, Kimi Code, and Antigravity. Run tasks in parallel, keep their project context together, and review the changes in one place.

Why use it instead of more terminal tabs?

Each task keeps its brief, agent, account, worktree, questions, and result together. You can see what needs attention and review related changes as one combined patch.

Do I need an AI subscription?

Bring a supported agent CLI and its provider account. Jackalope does not include model access. Subscription requirements, usage limits, and charges depend on your provider.

Does my code stay on my computer?

Repositories and task workspaces live on your computer. Agents may send code and context to their providers under your account settings. Connected tools have their own data policies.

Can I keep work and personal accounts separate?

Yes. Create named account profiles and choose project defaults. Continuations retain their selected profile. Antigravity profiles use Gemini API keys; its existing subscription login is shared. Profiles do not isolate local file access.

What happens after I join?

Verify your email to confirm your place and get your referral link. Each verified new signup through your link earns one day of priority. Approved members receive five Instant Access Passes to share. No payment is needed to join.

Can Jackalope choose an agent, model, and account?

Automatic tasks ask your configured default agent to choose an allowed agent, configured model, and permitted account. Jackalope validates project restrictions and tool compatibility, considers available capacity reports, and can hand off recognized quota failures to an eligible alternative while preserving the work. Explicit assignments remain available. Routing depends on your configured agents and available usage reports.

Which tools does Jackalope give agents?

Browse the MCP directory and choose project connections. Selected stdio and HTTP tools support on-demand discovery; direct connection support varies by agent. Built-in tools cover browser interaction, screenshots, accessibility audits, questions, and local verification. Windows desktop control adds accessibility snapshots, window captures, focus, clicks, typing, and scrolling after you select a window for that attempt. A theme-colored glow and status bar show when control is active. Escape or Stop revokes access; mouse or keyboard activity pauses it until you Resume. It operates your live desktop under existing OS permissions; macOS and Linux X11 window control is in native validation; Wayland remains unsupported.

How do different agents communicate and stay up to date?

Tasks share a project inventory with ownership, scopes, dependencies, and messages. Agents can send direct task messages or project broadcasts. Jackalope supplies a bounded briefing at launch and continuation, then attaches relevant updates to its own tool responses. Agents can check the durable inbox between those checkpoints. Delivery does not guarantee that an agent has read or acted on an update, and messages do not automatically wake agents, approve changes, or merge work.

Can I download Jackalope now?

Public downloads are not open yet. Join the waitlist and we’ll email you when access is ready.

Does an invitation include a download?

A valid direct invitation skips the waitlist after you verify your email, while the inviter has capacity. Your account shows downloads when a build is available.

Compare workspaces Explore the roadmap

Agent support & compatibility

Supported agents

Choose the right agent for each project. Keep the brief, attempts, changes, and review together when your tools change.

Codex

Your OpenAI coding agent, with project connections and reported usage.

Get the CLI

Claude Code

Use Claude Code with connected tools and answer its questions in Jackalope.

Get the CLI

Grok

Bring your Grok CLI sessions into the same task and review workflow.

Get the CLI

OpenCode

Choose your OpenCode provider and model, including available free or local options.

Get the CLI

Kimi Code

Use Kimi for tasks and routing, with separate accounts, model choices, and task approvals.

Setup guide

Antigravity

Run agy as a worker. Named accounts use Gemini API keys; subscription login is shared.

Get the CLI
Tasks & review · Session continuation · Usage when reported
Which features work with which agents?
AgentProject connectionsSign-in checks
Codexstdio · HTTPCLI status check
Claude Codestdio · HTTP · SSECLI status check
GrokOn-demand discoveryCLI account and model check
OpenCodestdio · HTTP · SSEProvider checks on launch
Kimi Codestdio · HTTP · SSECLI authentication; model access checked on launch
AntigravityOn-demand discoveryProvider checks on launch

Jackalope uses each agent’s installed CLI and permission rules. Connections configured inside a CLI remain subject to that CLI’s settings. Codex, Claude Code, OpenCode, and Kimi Code support direct project connections; all six support on-demand discovery. Kimi reports task token totals and membership quota. Antigravity reports subscription pools through read-only CLI commands. Usage coverage varies by account and CLI version. Profiles organize sign-ins; they are not a security sandbox.

Named profiles are available for supported agents. Antigravity profiles use Gemini API billing, separate from a Google subscription; multiple subscription logins are not isolated. Antigravity runs worker tasks; a verified tool-free interface is still needed for automatic routing and Ask Jackalope. Kimi supports both. Gemini CLI, Aider, and Goose have account setup only; task execution is still in development.

Model access and subscription requirements depend on your provider. See the roadmap for upcoming agent support.

Get early access.

We’ll email you when access opens.

See platform availability