Key takeaways
- Built-in MCP server enables agent-to-agent communication — unique among orchestrators
- Shipped six releases between February and April 2026 (v1.6 through v1.9), adding Autopilot, Personas, and artifact previews
- Relicensed from MIT to AGPL-3.0; still free and backed by Kochava (mobile analytics company)
FAQ
What is Skwad?
Skwad is an open-source macOS app that runs multiple AI coding agents with built-in MCP for inter-agent messaging and coordination.
How much does Skwad cost?
Skwad is free and open source under the AGPL-3.0 license (relicensed from MIT in early 2026).
Who competes with Skwad?
Supacode, Superset, Conductor, and Claude Squad offer similar multi-agent orchestration capabilities.
Executive Summary
Skwad is an open-source macOS app that runs multiple AI coding agents simultaneously with built-in MCP (Model Context Protocol) support for agent-to-agent messaging. From Kochava Studios, it differentiates through native agent coordination rather than just parallel execution.
Status (June 2026): Active. The project shipped six releases between February and April 2026 (v1.6.0 through v1.9.0, latest April 3, 2026) and launched a dedicated website, though the repo has been quiet since its last push on April 28, 2026. It also relicensed from MIT to AGPL-3.0.
| Attribute | Value |
|---|---|
| Company | Kochava Studios |
| Founded | 2026 |
| Funding | Internal (Kochava); no external funding publicly disclosed |
| Employees | 1 primary maintainer (+1 occasional Kochava contributor) |
| Headquarters | Sandpoint, Idaho |
Product Overview
Skwad positions itself as "your new, slightly revolutionary coding crew" — a macOS app where AI agents work together rather than just side-by-side. Each agent runs in its own embedded terminal, but the built-in MCP server lets them coordinate work themselves.
The app is built with Swift/SwiftUI and uses libghostty for GPU-accelerated terminal rendering (with SwiftTerm as fallback). It supports Claude Code, Codex, OpenCode, Gemini CLI, and GitHub Copilot out of the box, plus any custom CLI agent.
Releases through spring 2026 added an Autopilot mode (LLM-powered message classification and auto-responses), a Personas system for shaping agent coding styles, "The Bench" for saving and redeploying agent configurations, rich artifact previews (markdown and Mermaid diagrams), voice input, and desktop notifications.
Key Capabilities
| Capability | Description |
|---|---|
| Multi-Agent Management | Run multiple AI coding agents simultaneously |
| MCP Server | Built-in inter-agent messaging and coordination |
| GPU Terminals | libghostty-powered with SwiftTerm fallback |
| Git Integration | Worktree support, diff viewer, stage/commit panel |
| Activity Detection | See which agents are working or idle at a glance |
| Autopilot | LLM-powered message classification and auto-responses |
| Personas & The Bench | Shape agent coding styles; save and redeploy agent configurations |
| Artifacts | Markdown and Mermaid diagram previews from agent output |
Product Surfaces
| Surface | Description | Availability |
|---|---|---|
| macOS App | Native SwiftUI desktop app | Released (v1.9.0, April 2026) |
| GitHub Releases | Direct download of prebuilt builds | Available |
| Homebrew | brew install --cask skwad | Not yet available as of June 2026 |
Technical Architecture
Skwad uses a native macOS architecture with GPU-accelerated terminal rendering. The core differentiator is the built-in Hummingbird HTTP server that implements MCP, enabling agents to communicate with each other directly.
Key Technical Details
| Aspect | Detail |
|---|---|
| Deployment | Local macOS app |
| Terminal Engine | libghostty (GPU) / SwiftTerm (fallback) |
| Framework | SwiftUI, The Composable Architecture |
| MCP Server | Hummingbird HTTP |
| Open Source | Yes (AGPL-3.0; originally MIT) |
Dependencies
- libghostty — GPU-accelerated terminal from Ghostty project
- SwiftTerm — Fallback terminal emulation
- Hummingbird — HTTP server for MCP implementation
- swift-log — Logging framework
Strengths
- Agent-to-Agent Communication — The built-in MCP server is unique among orchestrators, enabling agents to coordinate work and hand off tasks without human mediation
- GPU-Accelerated Performance — libghostty provides native-quality terminal rendering that keeps up with fast-moving agent output
- Enterprise Backing — Kochava is an established company (mobile analytics/attribution), providing stability uncommon in indie tools
- Free and Open Source — Full app available at no cost under AGPL-3.0, with prebuilt downloads on GitHub Releases
- Rapid Release Cadence — Six releases shipped between February and April 2026, adding Autopilot, Personas, and artifact previews
- Native macOS Design — SwiftUI app follows platform conventions, integrates with macOS features
Cautions
- Minimal Adoption — 33 GitHub stars and 5 forks as of June 2026; effectively no community discussion on Hacker News or Reddit
- Development Has Cooled — No release since v1.9.0 (April 3, 2026) and no commits since April 28, 2026, after a rapid February–April cadence
- Near-Single Maintainer — Nicolas Bonamy remains the primary developer; one other Kochava contributor appeared in March 2026, so bus-factor risk persists
- AGPL-3.0 License — The relicense from MIT adds copyleft obligations that may deter commercial forks or embedding
- macOS 14.0+ Required — Requires Sonoma or later; no support for older macOS versions
- Build Complexity — Building from source requires the Zig toolchain or a prebuilt libghostty download, though prebuilt app downloads now exist
What Developers Say
As of June 2026, Skwad has essentially no public community footprint: a Hacker News search returns no submissions or discussion threads about it, and no substantive Reddit reviews surfaced. The most visible community activity is a maintainer talk ("Skwad: multi agent IDE") at an AI Tinkerers Chicago meetup. With no verbatim developer quotes available, adoption signals rest on repo metrics alone — 33 stars and 5 forks.
Pricing & Licensing
| Tier | Price | Includes |
|---|---|---|
| Open Source | $0 | Full app, AGPL-3.0 license |
Licensing model: AGPL-3.0 — free, copyleft (relicensed from MIT in early 2026)
Hidden costs: Requires AI coding CLI subscriptions (Claude Code, Codex, etc.) for actual agent functionality
Competitive Positioning
Direct Competitors
| Competitor | Differentiation |
|---|---|
| Supacode | Skwad has MCP for agent coordination; Supacode focuses on raw parallel scale (50+ agents) |
| Superset | Skwad emphasizes agent collaboration; Superset focuses on worktree management and notifications |
| Conductor | Skwad is open source and agent-agnostic; Conductor is closed-source from Melty Labs |
When to Choose Skwad Over Alternatives
- Choose Skwad when: You want agents to coordinate work themselves via MCP
- Choose Supacode when: You need maximum parallel scale (50+ agents) on macOS Tahoe
- Choose Superset when: You want polished UX with built-in diff viewer and PR workflow
Ideal Customer Profile
Best fit:
- Developers experimenting with multi-agent coordination patterns
- Open source enthusiasts who want to inspect and contribute
- Teams exploring MCP for agent communication
- macOS users comfortable with early-stage tools
Poor fit:
- Developers needing stable, battle-tested tools for production work
- Teams requiring Windows or Linux support
- Users who prefer GUI installers over build-from-source
- Anyone on macOS 13 or earlier
Viability Assessment
| Factor | Assessment |
|---|---|
| Financial Health | Moderate (Kochava-backed internal project; no external funding publicly disclosed) |
| Market Position | Niche (MCP-focused differentiator) |
| Innovation Pace | Slowing (six releases Feb–Apr 2026, then quiet since April 28, 2026) |
| Community/Ecosystem | Limited (33 stars, 5 forks, near-single maintainer, no public discussion) |
| Long-term Outlook | Uncertain (depends on Kochava commitment) |
Skwad is an interesting technical experiment from a stable company, and the spring 2026 burst of releases plus a dedicated website show real investment. But the MCP focus hasn't translated into adoption — stars only grew from 10 to 33 between February and June 2026 — and the repo has been quiet for about six weeks. Viability still depends on whether Kochava treats it as more than a side project. Nicolas Bonamy has built several open-source tools (including Witsy, a desktop AI assistant), suggesting sustained interest in the space.
Bottom Line
Skwad remains alive and worth watching for its unique MCP-based agent coordination approach, but it has not found an audience: after a fast February–April 2026 release run, activity paused and adoption stayed minimal (33 stars as of June 2026).
Recommended for: Developers exploring multi-agent coordination patterns who value open source and don't mind early-stage tools with tiny communities.
Not recommended for: Teams needing production-ready stability, cross-platform support, or tools with proven adoption; anyone whose use case conflicts with AGPL-3.0.
Outlook: The MCP-first approach is genuinely novel, and prebuilt releases plus a real website lowered the barrier to trying it. But with near-zero community traction and a six-week activity lull, the more likely path is that Skwad's core ideas influence other orchestrators while the project itself remains a Kochava side project.
Research by Ry Walker Research • methodology