Spec-Driven Development with AI Sub-Agents. An orchestrator + 9 specialized sub-agents for structured feature development. Zero dependencies. Pure Markdown. Works with Claude Code, OpenCode, Cursor, and more.
1,247
stars
73
commits
Shell
primary language
Mar 26, 2026
updated
README
Agent Teams Lite
This project has been deprecated in favor of
gentle-ai.Everything Agent Teams Lite provides — skills, orchestration, SDD workflow, skill registry — is now available through
gentle-aiwith a better installation experience, automatic updates, persistent memory via Engram, and support for 8 agents out of the box.
Migrating to gentle-ai
brew install gentleman-programming/tap/gentle-ai
gentle-ai install
That's it. gentle-ai detects your installed tools, installs all skills, configures MCP servers, injects the SDD orchestrator, and sets up Engram persistent memory — automatically.
What you get with gentle-ai that ATL didn't have
| Feature | Agent Teams Lite | gentle-ai |
|---|---|---|
| SDD orchestration (9 phases + judgment-day) | ✅ | ✅ |
| Skill registry + compact rules | ✅ | ✅ |
| branch-pr + issue-creation skills | ✅ | ✅ |
| Engram persistent memory (installed + configured) | ❌ | ✅ |
| Context7 documentation MCP | ❌ | ✅ |
| Persona injection | ❌ | ✅ |
| Automatic self-updates | ❌ | ✅ |
| Backup and rollback | ❌ | ✅ |
| Config sync and migration | ❌ | ✅ |
| GGA configuration | ❌ | ✅ |
| Permission management | ❌ | ✅ |
| TUI interactive installer | ❌ | ✅ |
Supported agents (all 8)
| Agent | Sub-agent support |
|---|---|
| Claude Code | Full |
| OpenCode | Full |
| Gemini CLI | Full |
| Codex | Full |
| Cursor | Inline |
| VS Code Copilot | Inline |
| Antigravity | Single-agent |
| Windsurf | Hybrid (Plan + Code mode) |
If you already have ATL installed
Run gentle-ai install — it will detect your existing setup and upgrade in place. Your skills, orchestrator config, and MCP servers will be migrated to the managed format with marker-based sections that support future sync and updates.
Why deprecate?
Maintaining feature parity across two distribution channels (shell scripts + Go binary) was unsustainable. Every change to a skill, orchestrator, or asset had to be replicated manually. With gentle-ai:
- One source of truth — assets are embedded in the binary, versioned with the release
- Atomic updates —
brew upgrade gentle-ai && gentle-ai syncrefreshes everything - No drift — marker-based injection replaces stale content instead of appending duplicates
- Cross-session memory — Engram protocol is injected into every agent's prompt automatically
This repo is archived
No new features or bug fixes will be made here. All development continues in gentle-ai.
The code remains available for reference. If you find value in the patterns described in the docs, they are all implemented (and improved) in gentle-ai.
License
Apache 2.0 — see LICENSE.
Contributors
72 commits
1 commits
Similar repos
Languages
Shell
62.4%
PowerShell
37.6%
Spec-Driven Development with AI Sub-Agents. An orchestrator + 9 specialized sub-agents for structured feature development. Zero dependencies. Pure Markdown. Works with Claude Code, OpenCode, Cursor, and more.
1,247
stars
73
commits
Shell
primary language
Mar 26, 2026
updated
README
Agent Teams Lite
This project has been deprecated in favor of
gentle-ai.Everything Agent Teams Lite provides — skills, orchestration, SDD workflow, skill registry — is now available through
gentle-aiwith a better installation experience, automatic updates, persistent memory via Engram, and support for 8 agents out of the box.
Migrating to gentle-ai
brew install gentleman-programming/tap/gentle-ai
gentle-ai install
That's it. gentle-ai detects your installed tools, installs all skills, configures MCP servers, injects the SDD orchestrator, and sets up Engram persistent memory — automatically.
What you get with gentle-ai that ATL didn't have
| Feature | Agent Teams Lite | gentle-ai |
|---|---|---|
| SDD orchestration (9 phases + judgment-day) | ✅ | ✅ |
| Skill registry + compact rules | ✅ | ✅ |
| branch-pr + issue-creation skills | ✅ | ✅ |
| Engram persistent memory (installed + configured) | ❌ | ✅ |
| Context7 documentation MCP | ❌ | ✅ |
| Persona injection | ❌ | ✅ |
| Automatic self-updates | ❌ | ✅ |
| Backup and rollback | ❌ | ✅ |
| Config sync and migration | ❌ | ✅ |
| GGA configuration | ❌ | ✅ |
| Permission management | ❌ | ✅ |
| TUI interactive installer | ❌ | ✅ |
Supported agents (all 8)
| Agent | Sub-agent support |
|---|---|
| Claude Code | Full |
| OpenCode | Full |
| Gemini CLI | Full |
| Codex | Full |
| Cursor | Inline |
| VS Code Copilot | Inline |
| Antigravity | Single-agent |
| Windsurf | Hybrid (Plan + Code mode) |
If you already have ATL installed
Run gentle-ai install — it will detect your existing setup and upgrade in place. Your skills, orchestrator config, and MCP servers will be migrated to the managed format with marker-based sections that support future sync and updates.
Why deprecate?
Maintaining feature parity across two distribution channels (shell scripts + Go binary) was unsustainable. Every change to a skill, orchestrator, or asset had to be replicated manually. With gentle-ai:
- One source of truth — assets are embedded in the binary, versioned with the release
- Atomic updates —
brew upgrade gentle-ai && gentle-ai syncrefreshes everything - No drift — marker-based injection replaces stale content instead of appending duplicates
- Cross-session memory — Engram protocol is injected into every agent's prompt automatically
This repo is archived
No new features or bug fixes will be made here. All development continues in gentle-ai.
The code remains available for reference. If you find value in the patterns described in the docs, they are all implemented (and improved) in gentle-ai.
License
Apache 2.0 — see LICENSE.
Similar repos
Contributors
72 commits
1 commits
Languages
Shell
62.4%
PowerShell
37.6%