Platforms

Every platform has its own way of doing things. Here's how teamrc handles each one.

Overview

teamrc auto-detects which platforms you have installed and generates the right files for all of them. You don't need to think about the differences. That's the whole point. If you want to target specific platforms only, use the --platform flag.

File naming convention

All teamrc-generated files are prefixed with trc- . This makes them easy to identify and ensures they don't collide with files you create manually. When teamrc syncs, it only touches files with this prefix.

Claude Code

Detection: ~/.claude/ directory exists. Supports both project and global scope.

Generated files

agents → .claude/agents/trc-{name}.md
always-apply skills → .claude/rules/trc-{id}.md
glob skills → .claude/rules/trc-{id}.md
on-demand skills → .claude/skills/trc-{id}/SKILL.md
built-in commands → .claude/skills/trc-{save-knowledge,save-core,knowledge,status}/
knowledge → .teamrc/knowledge-{slug}.md
team context → updated in CLAUDE.md

Agent files use YAML frontmatter with name, description, and a skill list. Per-agent skills are listed in the frontmatter so Claude Code natively routes them. teamrc also appends a team context block to CLAUDE.md describing the team and its members. Built-in slash commands ( /trc-save-knowledge , /trc-knowledge , etc.) are installed as on-demand skills for interacting with team knowledge mid-session.

Cursor

Detection: .cursor/ directory in the project. Project scope only.

Generated files

agents → .cursor/agents/trc-{name}.md
always-apply/glob skills → .cursor/rules/trc-{id}.mdc
on-demand skills → .cursor/skills/trc-{id}/SKILL.md
built-in commands → .cursor/skills/trc-{save-knowledge,save-core,knowledge,status}/
knowledge → .teamrc/knowledge-{slug}.md
routing → .cursor/AGENTS.md

Rule files use the .mdc format (Markdown Config) with YAML headers for description, globs, and alwaysApply. Team routing information is written to .cursor/AGENTS.md .

Codex

Detection: .codex/ directory. Project scope only.

Generated files

agents → .codex/agents/trc-{name}.toml
team config → .codex/config.toml (multi-agent section)
routing → AGENTS.md
on-demand skills → .agents/skills/trc-{id}/SKILL.md
knowledge → .teamrc/knowledge-{slug}.md

Codex uses TOML-based configuration. Agent instructions are stored as triple-quoted strings. Multi-agent mode is automatically enabled in config.toml . Always-apply and glob-targeted skills are written into AGENTS.md .

Gemini

Detection: .gemini/ directory. Supports both project and global scope.

Generated files

agents → .gemini/agents/trc-{name}.md
skills → .agents/skills/trc-{id}/SKILL.md
knowledge → .teamrc/knowledge-{slug}.md
team context → GEMINI.md

Agent files use YAML frontmatter with a team header section. Skills are written to both .agents/skills/ and .agent/skills/ (or the global config directory) for compatibility. Always-apply skills are inlined in GEMINI.md .

OpenClaw

Detection: .agents/ directory. Supports both project and global scope.

Generated files

agents → .agents/agents/trc-{name}.md
skills → .agents/skills/trc-{id}/SKILL.md
knowledge → ~/.openclaw/knowledge-{slug}.md
routing → AGENTS.md

Uses the native OpenHands format with YAML frontmatter. Per-agent skill assignments are listed in the agent's frontmatter. Team routing is written to AGENTS.md .

Coming soon

Adapters for these platforms are planned but not yet implemented: GitHub Copilot, Amazon Q, Windsurf, and Cline.