Agent Skills for three.dev,
bundled with the three.dev MCP server at https://mcp.three.dev/mcp.
Paste this into your coding agent:
Install github.com/round3ai/skills and run its three-dev-setup skill in this repository.
# Claude Code
claude plugin marketplace add round3ai/skills
claude plugin install three-dev@three-dev
# Codex
codex plugin marketplace add round3ai/skills
codex plugin install three-dev
# Any other agent: cursor, github-copilot, kiro-cli, cline, ...
npx skills add round3ai/skills -y --agent <agent>The Claude Code and Codex plugins also set up the three.dev MCP server. Sign in
when your client asks, or run /mcp in Claude Code.
If you installed with npx skills, add the MCP server to your client's config:
| Client | Config |
|---|---|
| Cursor | .cursor/mcp.json: {"mcpServers": {"three-dev": {"url": "https://mcp.three.dev/mcp"}}} |
| VS Code | .vscode/mcp.json: {"servers": {"three-dev": {"type": "http", "url": "https://mcp.three.dev/mcp"}}} |
| Kiro | .kiro/settings/mcp.json: {"mcpServers": {"three-dev": {"url": "https://mcp.three.dev/mcp"}}} |
| Cline | MCP Servers panel: the URL above, transport Streamable HTTP |
Then paste this into your coding agent:
Run the three-dev-setup skill in this repository.
Update with claude plugin update three-dev or npx skills update.
| Skill | What it does |
|---|---|
three-dev-experiments |
Runs an offline experiment in three.dev to test a prompt, model, or reasoning change on recorded production traffic, and reads the results. |
three-dev-failure-modes |
Investigates production quality problems in an LLM feature that three.dev already records. |
three-dev-quality-metrics-setup |
Sets up quality metrics for an LLM feature that three.dev already records. |
three-dev-setup |
Adds three.dev to a codebase so its LLM calls are recorded. |
The skills are published here automatically and also served by the three.dev
MCP server. Changes to skills/ or the table above are overwritten on the next
publish, so open an issue instead.