Skip to content

START_HERE.md

Latest commit

 

History

History
73 lines (51 loc) · 4.09 KB

START_HERE.md

File metadata and controls

73 lines (51 loc) · 4.09 KB
Raw

Get started with rdrop for Raindrop.io

Install rdrop, connect your Raindrop account, then use your shell or agent. An existing Codex/Claude agent already supplies the model: you do not need another AI account, Laya, Executor or ACP to get started.

1. Install

Choose one:

brew install machineyearningengineer/tap/rdrop
# Or Node 22+, with optional dependencies enabled:
npm install --global @machineyearn/raindrop-cli@0.2.1
rdrop version

Without a global npm install, replace rdrop in the commands below with npx --yes @machineyearn/raindrop-cli@0.2.1. Native downloads, source builds and checksums.

2. Connect Raindrop

Create a token in Raindrop's integration settings. On macOS/Linux, prepare a private configuration file:

mkdir -p "$HOME/.config/rdrop"
chmod 700 "$HOME/.config/rdrop"
touch "$HOME/.config/rdrop/env"
chmod 600 "$HOME/.config/rdrop/env"

Open that file in your text editor and add RAINDROP_TOKEN=your_token_here, replacing the placeholder privately. On Windows, the default file is %USERPROFILE%\.config\rdrop\env; keep it accessible only to your account. Then:

rdrop doctor
rdrop cleanup report --json

The doctor checks configuration and account access without printing the token. The report reads your library; it does not change bookmarks. Environment variables override configuration files. If you use XDG_CONFIG_HOME or another location, run rdrop config paths to see the paths this installation reads. Configuration reference.

3. Use your agent or shell

Copy the setup and organizing prompt from the README into your agent. It tells the agent to check setup, use existing tools, start with a report and small sample, then show exact changes before applying an approved plan.

For Codex or Claude, install the shared plugin. Other local clients can use the stdio MCP configuration. The default MCP server reads and plans. Installing the plugin does not enable writes.

Or start directly:

rdrop collections --tree
rdrop search 'type:article' --json
rdrop cleanup report --json

Your existing tag names and collection paths are the starting vocabulary. Review a bounded sample before extending organization to the rest of the library. Reviewed planning and recovery.

Add capabilities when you need them

Need Use
Raindrop's semantic search, extracted content or AI audits Add --official to mcp serve or agent; upstream plan limits apply
Run a tool-capable model without an external agent rdrop agent with a hosted or local model
Fixed categories with visible uncertainty classify with configured Jev, Laya or Clef
Local model inference Qwen/Laya setup and verification
Portable tools in Executor Executor setup
Give this MCP server to an existing ACP agent ACP configuration

These are optional routes. Keep the first setup small, then choose a provider or host for a concrete job.

If setup fails

  • Command not found: restart the terminal, check the npm/Homebrew bin directory is on PATH, or use npx. GUI agents may need an absolute executable path.
  • Missing token or HTTP 401: inspect rdrop config paths and rdrop config show, then replace the token in the private file. Do not paste it into chat.
  • Missing npm platform package: reinstall with npm install --global --include=optional @machineyearn/raindrop-cli@0.2.1, or use a native archive.
  • No agent tools: reload/start a new client session and check its MCP connection. Browser-only clients require a hosted HTTP service; this local server uses stdio.
  • Official/model provider unavailable: check that provider's configuration and account limits. The shell CLI and normal REST reads do not depend on that provider.