Works in
VS CodeVS CodeCursorCursorAntigravity BetaAntigravity Beta

The IDE for Agents.
Run Claude, Codex, and Gemini side-by-side as editor tabs.

Specify your Mix of Agents in one command. Swarmify orchestrates Claude, Codex, Gemini, and Cursor in parallel - each with its own context window. Describe your task, get a distribution plan, approve it, and watch your swarm deliver.

CLI agents
Get Swarmify
Agents run as editor tabs, not hidden terminals
my-project - Cursor
> Update agents.ts to add OpenCode support
Adding OpenCode to AGENT_TYPES enum and spawn modes. Spawning:
Swarm:Spawn task="update-interfaces" agent=codex
agent_id: 7a4f82c1
Swarm:Spawn task="update-constants" agent=gemini
agent_id: 3b9e15d8
Polling status...
7a4f82c1: rush tool file_editor src/agents.ts
3b9e15d8: rush tool file_editor src/constants.ts
AgentsChanges5
Main
+4~1
Diff Preview
- export const AGENT_TYPES = ['Claude', 'Codex', 'Gemini', 'Cursor'];
+ export const AGENT_TYPES = ['Claude', 'Codex', 'Gemini', 'Cursor', 'OpenCode'];
+ export const SPAWN_MODES = ['plan', 'edit', 'ralph'] as const;
GitHubCommit
feat: add OpenCode agent and Ralph mode support

How it works

CLI agents run as editor tabs. Swarmify MCP enables them to spawn sub-agents for parallel work.

L1
CLI coding agents
ClaudeClaude
CodexCodex
GeminiGemini
CursorCursor
L2
Task delegation
Build auth system
CodexOAuth
Google
GitHub
GeminiSessions
Redis
Tokens
ClaudeRoutes
Login
Logout
L3
IDE integration
VS CodeVS Code
CursorCursor

Agents Extension

Editor tabs, labeling, and task systems that keep agent sessions visible and organized.

Prompt library and task launching

Save prompts and launch tasks with consistent context. Promote any output into a new task without leaving the editor.

Search and reuse prompts across projects.
Kick off complex tasks with one shortcut.
Labels flow into tab titles for quick scanning.
Swarmify - Agents
Search prompts...
research plan implement
Build a plan, split tasks, then execute with scoped CLI agents and review diffs.
debug consensus
Launch multiple agents to confirm root cause and compare findings.

Markdown and todo workflows

Render agent markdown output like a real document, check todos, and convert them into tasks.

Readable markdown with tables and inline code.
Todos launch tasks directly into a new tab.
Diff previews stay attached to the agent session.
README.md
Landing Page Tasks

This doc tracks the next batch of updates. Review diffs as agents work and ship changes in parallel.

Highlights
  • Hero copy aligned to orchestration
  • Prompt library visible by default
  • Task list hooks for instant delegation

Agents MCP

MCP server for multi-agent orchestration. Swarm mode lets you spawn and coordinate multiple CLI agents in parallel under one orchestrator.

$npx @swarmify/agents-mcp

Orchestrator control

One orchestrator coordinates multiple CLI agents with scoped tasks and shared context.

Swarmify MCP

Use the MCP server for structured task spawning and consistent output formatting.

Installation

Install the extension once. All agent sessions appear as editor tabs.

Context alignment

AGENTS.md, CLAUDE.md, and GEMINI.md are kept in sync with symlinks so every agent sees the same ground truth.

Native subagent gap

Many CLI agents still lack native subagent spawning. Swarmify bridges this gap.

Async subagents

Spawned agents run asynchronously so the orchestrator can keep working. Check back later to review results.

Parallel execution that stays visible

Subagents run in parallel while the orchestrator continues working. When they finish, their tabs show diffs and markdown output for review.

OrchestratorRunning
Subagent ARunning
Subagent BRunning

Agents CLI

Unified CLI for AI coding agents. Manage commands, MCP servers, skills, and CLI versions across Claude, Codex, Gemini, and more.

$npm install -g @swarmify/agents-cli

Commands

Install slash commands to all agents at once. Auto-converts Markdown to TOML for Gemini.

Skills

Install Agent Skills (SKILL.md + rules/) from repos. Share capabilities across all your agents.

MCP Servers

Register and unregister MCP servers across Claude, Codex, and Gemini in one command.

CLI Versions

Pin versions, upgrade all agents, and sync CLI configurations across machines.

Hooks

Sync hooks and their data files across machines. Keep all agents configured identically.

Sync

Push config to GitHub, pull on new machines. Set up a fresh environment in seconds.

Quick Start

$agents status
$agents commands add gh:user/commands
$agents mcp add swarm "npx @swarmify/agents-mcp"
$agents pull gh:username/.agents

Supported Agents

ClaudeMCP, Hooks, Skills
CodexMCP, Skills
GeminiMCP, Hooks, Skills
CursorConfig
OpenCodeConfig
TraeConfig

Why multiple CLI agents

Different CLI agents are stronger at different tasks. Swarmify lets the orchestrator run them in parallel so you get the best model for each slice of work.

Codex

Highly surgical for large codebases and precise edits across many files.

Claude Opus 4.5

Very capable for complex reasoning, planning, and long chain decisions.

Gemini and Cursor

Strong for targeted refactors, fast iteration, and complementary perspectives.

Why CLI agents in an IDE

SSH support lets IDEs like VS Code run anywhere. Swarmify gives you visibility into agent work inside the IDE instead of running agents blindly in a terminal.

Terminal only

Limits
  • Minimal visibility into agent edits
  • Hard to compare multiple outputs
  • Context switching between panes
  • Markdown, diffs, and previews are hidden

IDE with Swarmify

Swarmify
  • Live diffs and markdown in editor tabs
  • Parallel agent tabs with clear labels
  • Split view of code and agent output
  • SSH support means the IDE can run anywhere

Integration guide

Pick your IDE and get started in minutes.

VS Code

Install the Swarmify extension, then run agents in VS Code terminal tabs.

Cursor

Install the extension, then launch agent sessions as editor tabs.

VS Code over SSH

Use Remote SSH to run on any machine while keeping agents visible in the local IDE.

Developer experience

Productivity features that make multi-agent workflows feel native.

New

Session persistence

All agent sessions are saved to disk automatically. Crash, close VS Code, or restart your machine. Pick up exactly where you left off.

Zero data loss, automatic recovery
New

AI git commits

Generate commit messages that match your style. One shortcut to stage, commit, and push with a message that fits your repo conventions.

Cmd+Shift+G

Keyboard-first workflow

12+ shortcuts for spawning agents, switching tabs, labeling sessions, and launching prompts. Stay in flow without reaching for the mouse.

Cmd+Shift+NCmd+Shift+XCmd+Shift+L

TODO.md discovery

Drop a TODO.md in your workspace and Swarmify finds it. Spawn agents directly from task items without copy-pasting prompts.

Works with checkbox format and markdown headers

Common workflows

Multi-agent execution with clear orchestration and IDE visibility.

Orchestrate parallel work from one prompt

Swarmify turns a single orchestrator request into multiple focused CLI agent tasks. Each agent tackles a clear slice and reports back in tabs.

  1. Orchestrator splits a plan into scoped sub tasks
  2. Agents work in parallel across files and services
  3. Diffs and markdown stay visible in the editor
Best results come from a deliberate research plan implement flow
Example
Implement Stripe checkout with tests and docs
ClaudeClaude- Orchestrator plan
CodexCodex- Tests
FILEcheckout.ts
ClaudeClaude Code
> Implement Stripe checkout with tests and docs
Planning scoped tasks for parallel execution...
Reading: src/billing/stripe.ts
Editing: src/api/checkout.ts
Task list created for subagents.

Validate findings with multiple agents

Different CLI agents are stronger at different tasks. Swarmify puts their findings side by side so the orchestrator can converge on the right fix.

  1. One agent maps the failure path
  2. Another inspects edge cases and logs
  3. Orchestrator merges the final change set
Example
Trace why guest checkout fails in production
ClaudeClaude- Orchestrator plan
CodexCodex- Tests
FILEcheckout.ts
ClaudeClaude Code
> Trace why guest checkout fails in production
Planning scoped tasks for parallel execution...
Reading: src/billing/stripe.ts
Editing: src/api/checkout.ts
Task list created for subagents.

Used by engineers from

Google
TikTok

FAQ

Do agents step on each other's work?

The /swarm command tells the orchestrator to split work carefully and assign clear scopes. It is like giving instructions to human coworkers, each with a defined task and boundary.

How do I review results from subagents?

Each subagent reports back in its own tab with markdown and diffs. The orchestrator can merge the final change set after review.

What about context drift?

Swarmify manages AGENTS.md, CLAUDE.md, and GEMINI.md with symlinks so all agents share the same context source.

Try Swarmify

Run multiple CLI agents in parallel, review every diff, and ship faster.

Get Swarmify for VS Code