Documentation

How Reddy fits into your AI coding workflow, and exactly what it can and can't touch.

Getting started

Install Reddy once into the tool you build with (see Get Reddy), then in each app ask your assistant to connect its services, or run:

$ npx -y <your-npm-scope>/cli@0.1.0 init

Reddy works out what the app needs, gives you one setup link, verifies each provider with live checks, and writes the configuration into your project after you approve. Installing Reddy and registering a project are separate: one installation serves all your projects.

CLI commands

reddy install
Add Reddy to Claude Code, Cursor or the terminal. Shows the plan first; agents must pass --yes only after you agree.
reddy init
Register the current project: detects services from dependency and variable names, writes reddy.yaml, returns a setup link.
reddy status
What's connected and verified, what's missing, whether your env file matches, and the next step. --json for agents.
reddy connect
Create a 30-minute setup link, bound to your account and this project.
reddy sync
Write approved configuration to your local env file. Secrets need your browser approval. --no-wait and --continue for agents.
reddy verify
Re-run live checks. --local checks the values in your own env file without printing them.
reddy doctor
Check every layer: runtime, sign-in, each AI tool's MCP server, project, Git safety. --fix for safe repairs.
reddy uninstall
Remove Reddy from a tool (--tool) or everywhere (--all) and revoke its sign-in.

MCP tools

One MCP server serves every MCP-capable tool. It runs on your computer, so it can read the project and write approved configuration, and it ships instructions that tell the assistant when to use Reddy and never to ask you for keys in chat.

reddy_get_account_status
Signed in? As whom? Up to date?
reddy_sign_in
Start or finish sign-in for this tool (the user approves in the browser).
reddy_detect_project
Registered, unregistered, copied from elsewhere, or not a project.
reddy_register_project
Register the project and get a setup link.
reddy_analyze_dependencies
Which services the code appears to need, with evidence. Local only.
reddy_list_required_services
Services and chosen providers from reddy.yaml.
reddy_get_connection_status
Full status, including whether the local env file matches.
reddy_discover_provider
What Reddy can do for a provider. For providers it doesn't know, researches the official docs into a cited setup guide.
reddy_generate_setup_plan
Plain-language plan of what's left and who does it. Can add a discovered provider to an empty slot.
reddy_create_setup_session
One setup link for everything that's missing.
reddy_get_setup_progress
Is setup done? Can wait up to 50 seconds.
reddy_request_environment_sync
Write approved configuration into the project; returns an approval link when secrets are involved.
reddy_verify_integration
Re-run live checks.
reddy_get_safe_project_config
Variable names, browser-safe or server-only. Never values.
reddy_get_provider_preferences
Providers used before and accounts already connected.
reddy_get_integration_history
Recent setup, verification and sync activity.

reddy.yaml

A small file in your repository that records which services the app uses and which environment variable names it reads. Commit it: it never contains secret values, and Reddy refuses to write or read one that does. Because a repository can come from anywhere, Reddy validates it strictly and never takes a server address from it.

Security model

  • Credentials are encrypted with a per-value key and bound to one project, so one app can't read another's.
  • Every installation (terminal, Claude Code, Cursor) has its own sign-in that you approve in the browser and can revoke on its own.
  • Installations can request setup links and syncs. Secret values only reach your computer after you approve that specific sync, once.
  • No tool result, CLI output, log or manifest contains a secret value. Agents see variable names and status only.
  • Production secrets are never written to local files.
  • Reddy only writes .env files in the project root, refuses files tracked by Git, and never overwrites a value without asking. It keeps a copy of the old file only until the new one is confirmed, and never one with a replaced secret.

Uninstalling

Remove Reddy from one tool, or everywhere. Your code, env files, reddy.yaml and provider accounts are never touched.

$ npx -y <your-npm-scope>/cli@0.1.0 uninstall --tool cursor
$ npx -y <your-npm-scope>/cli@0.1.0 uninstall --all

You can also disconnect any installation from Settings, under AI Tools & Installations. Deleting your Reddy account is a separate step.