Skip to content

Divinci Agent Skill for Claude Code

Copy page

The Divinci skill is a packaged set of instructions (Agent Skills) that teaches a coding agent how to build with the Divinci platform — the divinci CLI, the browser and Node.js SDKs, the hosted MCP server, and the raw REST API. With it installed, your agent picks the right surface, uses correct auth patterns, and avoids the common footguns (embedding server keys in browsers, holding uploads open past edge timeouts, mixing embedding models in one RAG vector) without you pasting docs into every prompt.

divinci/
SKILL.md # concepts, auth, "which surface do I use?"
references/
cli.md # divinci CLI command reference + scripting patterns
client-sdk.md # @divinci-ai/client — headless browser chat
server-sdk.md # @divinci-ai/server — backend automation
mcp.md # hosted MCP server + @divinci-ai/mcp

The skill uses progressive disclosure: agents load the small SKILL.md first and pull reference files only when a task needs them, so it costs almost nothing in context until it’s relevant.

  1. Download the package: divinci-skill.zip

  2. Unzip it into your skills directory:

    Terminal window
    unzip divinci-skill.zip -d .claude/skills/
  3. That’s it. The skill activates automatically whenever a task involves Divinci, or invoke it explicitly:

    /divinci

The skill teaches the agent how to use Divinci; the hosted MCP server gives it live access to your account (search knowledge, upload files, inspect releases). They compose well:

Terminal window
claude mcp add divinci --transport http https://mcp.divinci.app/mcp

Once installed, prompts like these just work:

  • “Add a streaming Divinci chat panel to this React app for release rel_….”
  • “Script a CI job that uploads docs/**/*.md to our RAG vector and fails if any file lands in failedChunkSources.”
  • “Create a LightRAG target for our product docs and wire the release to it.”
  • “Expose this release as an MCP server and connect Claude Desktop to it.”