Skip to main content

Use Distri from your coding agent

Distri ships machine-readable instructions so your coding agent, Claude, OpenAI, Cursor, or anything that reads a rules file, can create, run, and deploy Distri agents naturally. Everything is a single curl away.

Install the skill (Claude Code / Claude Agent SDK)​

curl -fsSL https://distri.dev/skills/install.sh | sh

This drops a focused distri skill into .claude/skills/distri/. Claude picks it up automatically. The skill covers the CLI, the agent.md format, running and serving, pushing to the cloud, reading traces, and embedding.

Install somewhere else:

curl -fsSL https://distri.dev/skills/install.sh | sh -s -- path/to/skills/distri

For any agent (no skill system needed)​

Point your rules file or system prompt at the skill, or paste the full reference:

curl -fsSL https://distri.dev/skills/distri/SKILL.md   # the skill, self-contained
curl -fsSL https://distri.dev/llms-full.txt            # complete API + CLI + examples
curl -fsSL https://distri.dev/llms.txt                 # short index of docs
  • Cursor, add the SKILL.md content to .cursor/rules (or .cursorrules).
  • OpenAI / Codex / others, include SKILL.md in the system prompt, or fetch llms-full.txt when the model needs the full API.
  • AGENTS.md, drop the SKILL.md content into your repo's AGENTS.md.

What's available​

ResourceURL
Skill installerhttps://distri.dev/skills/install.sh
Skill entryhttps://distri.dev/skills/distri/SKILL.md
Agent formathttps://distri.dev/skills/distri/references/agents.md
CLI referencehttps://distri.dev/skills/distri/references/cli.md
Traceshttps://distri.dev/skills/distri/references/traces.md
Skills & workflowshttps://distri.dev/skills/distri/references/workflows-and-skills.md
Embedding & SDKhttps://distri.dev/skills/distri/references/embedding.md
Skills indexhttps://distri.dev/skills.txt
Full text referencehttps://distri.dev/llms-full.txt
Docs index (llms.txt)https://distri.dev/llms.txt

Best practice: one skill, many references​

Distri ships a single distri skill (not a dozen fragmented ones) with reference files bundled beside it. That keeps activation reliable, your agent loads the one skill when a task mentions Distri, then reads only the reference it needs (agents, CLI, traces, workflows, embedding). This mirrors the recommended structure for Claude Code / Claude Agent SDK skills.