What AI Agents Does gstack Support? (Codex, Cursor, OpenClaw & More)
gstack supports 10 AI coding agents including OpenAI Codex CLI, Cursor, and OpenClaw through the SKILL.md standard, with auto-detection and host-specific skill bundles.
The garrytan/gstack repository is built to run on any AI-coding host that implements the SKILL.md standard. If you're wondering what AI agents gstack supports, the framework ships ready-to-use skill bundles for ten agents, with Claude configured as the default host.
Full List of Supported AI Agents
According to the README.md and CONTRIBUTING.md files, gstack officially supports the following ten agents. Each requires a specific setup flag and installs skills to a dedicated directory:
| Agent | Setup Flag | Install Path for Generated Skills |
|---|---|---|
| Claude (primary) | (default) | ~/.claude/skills/gstack-*/ |
| OpenAI Codex CLI | --host codex |
~/.codex/skills/gstack-*/ |
| Cursor | --host cursor |
~/.cursor/skills/gstack-*/ |
| OpenClaw | --host openclaw |
~/.openclaw/skills/gstack-*/ |
| Factory | --host factory |
~/.factory/skills/gstack-*/ |
| Kiro | --host kiro |
~/.kiro/skills/gstack-*/ |
| OpenCode | --host opencode |
~/.config/opencode/skills/gstack-*/ |
| Slate | --host slate |
~/.slate/skills/gstack-*/ |
| Hermes | --host hermes |
~/.hermes/skills/gstack-*/ |
| GBrain (mod) | --host gbrain |
~/.gbrain/skills/gstack-*/ |
How gstack Detects and Configures AI Agents
The ./setup script determines which agent to target through three mechanisms:
-
Auto-detection: The script scans your
$PATHfor the binaries of supported agents and automatically selects the first one found. -
Explicit flag: Override auto-detection by passing
./setup --host <name>, using the flags shown in the table above. -
Host-specific generation: The code generator reads TypeScript configurations from
hosts/*.ts(e.g.,hosts/codex.ts,hosts/cursor.ts) to emit tailoredSKILL.mdbundles. These files define agent-specific front-matter, path conventions, and command aliases required by each platform.
OpenClaw Integration Details
Unlike other agents that run gstack's complete skill set natively, OpenClaw operates through a partial import model. OpenClaw imports the gstack-lite and gstack-full plan files alongside four native methodology skills: office-hours, ceo-review, investigate, and retro.
The dispatch routing rules are defined in openclaw/agents-gstack-section.md, which OpenClaw copies into its own AGENTS.md file. This architecture allows users to invoke any gstack skill from within an OpenClaw session while maintaining OpenClaw's native runtime environment.
Installing gstack for Codex, Cursor, and OpenClaw
To install skill bundles for the three most requested agents:
# Clone the repository
git clone --depth 1 https://github.com/garrytan/gstack.git ~/gstack
cd ~/gstack
# Install for Codex CLI
./setup --host codex
# Install for Cursor
./setup --host cursor
# Install for OpenClaw (includes plan files and native skills)
./setup --host openclaw
Each command writes the generated SKILL.md files to the appropriate home-directory path shown in the support table.
Running gstack Skills Across Different Agents
After installation, invoke skills using each agent's native command interface:
# From a Codex-CLI session (after codex login):
/office-hours
# From a Cursor terminal session:
/plan-eng-review
# From an OpenClaw chat session:
/gstack-openclaw-investigate
Extending Support to New AI Agents
To add support for an AI agent not currently listed, create a new host configuration file under hosts/myagent.ts following the structure documented in docs/ADDING_A_HOST.md. The architecture requires only this single TypeScript file—no changes to the core logic are necessary. Regenerate all skill bundles with bun run gen:skill-docs --host myagent to produce the new host-specific documentation.
Summary
- gstack supports 10 AI agents including OpenAI Codex CLI, Cursor, and OpenClaw through standardized SKILL.md bundles.
- Auto-detection via
./setupscans$PATHfor installed agents, or use--hostflags for explicit targeting. - Host-specific configurations in
hosts/*.tsdrive code generation for each platform's unique requirements. - OpenClaw uniquely imports plan files and four native skills via dispatch rules in
openclaw/agents-gstack-section.md. - Modular architecture allows extension to new agents by adding a single TypeScript file under
hosts/.
Frequently Asked Questions
Does gstack support OpenAI Codex CLI?
Yes. gstack fully supports OpenAI Codex CLI through the --host codex flag. The installer generates a tailored SKILL.md bundle and writes it to ~/.codex/skills/gstack-*/, enabling slash commands like /office-hours directly within Codex sessions.
How do I install gstack for Cursor IDE?
Run ./setup --host cursor from the repository root after cloning. This detects Cursor's environment and installs the skill bundle to ~/.cursor/skills/gstack-*/. Once complete, you can invoke gstack skills from Cursor's built-in terminal using standard slash commands.
What makes OpenClaw integration different from other agents?
OpenClaw does not host the full gstack skill set directly. Instead, it imports gstack-lite and gstack-full plan files plus four methodology skills (office-hours, ceo-review, investigate, retro). The integration uses dispatch routing rules from openclaw/agents-gstack-section.md, which OpenClaw merges into its AGENTS.md to enable skill invocation without native hosting.
Can I use gstack with AI agents not on the official list?
Yes. Any AI-coding host implementing the SKILL.md standard can be supported. Create a configuration file in hosts/myagent.ts following the template in docs/ADDING_A_HOST.md, then regenerate skills with bun run gen:skill-docs --host myagent. The system requires only this single TypeScript configuration file to generate complete support.
Have a question about this repo?
These articles cover the highlights, but your codebase questions are specific. Give your agent direct access to the source. Share this with your agent to get started:
curl -s "https://instagit.com/install.md" Maintain an open-source project? Get it listed too →