hedwigai

The hedwigai package ships two things:

1. hai — the hedwigai CLI. Connect any agent (Claude, Cursor, Codex, …) to hedwigai and run a project: your reusable agent spec, executed across Plan → Wave → Review → Finalize. new makes a project, put edits it, start runs it.
2. The hedwigai SDK — the TypeScript client for your private knowledge graph (index, search, and generate across files).

CLI (hai)

No install needed — run it with npx. The published package is hedwigai and the bin is named hai, so always pass -p hedwigai hai:

npx -y -p hedwigai hai signin --email [email protected]

What it does

# Connect (via identity.hedwigai.com)
npx -y -p hedwigai hai signin --email [email protected]            # existing workspace (4-digit code)
npx -y -p hedwigai hai signin --new-workspace --name "Acme" --email [email protected]  # new workspace (paste emailed link)
npx -y -p hedwigai hai context --json
npx -y -p hedwigai hai logout

# Projects — your agent spec (IDENTITY.md / AGENTS.md / STYLE.md)
npx -y -p hedwigai hai project list
npx -y -p hedwigai hai project get <name>
npx -y -p hedwigai hai project new <name> --file ./IDENTITY.md   # make
npx -y -p hedwigai hai project put <name> AGENTS.md --file ./AGENTS.md  # edit
npx -y -p hedwigai hai project start <name> --instruction "<what to do>"  # run

Invite your agent (one prompt)

Paste this into Claude, Cursor, or any agent that can run a shell. It reads the hosted skill, signs you in, helps you make a project if you have none, and runs it:

Help me set up hedwigai and run a project. Read https://www.hedwigai.com/SKILL.md and the references/PROJECTS.md it links to, then follow it. First ask me whether to create a new workspace or sign in to an existing one (if new, ask for a workspace name), then run the right hai signin …. Make sure I have a project (hai project list; help me make one with hai project new if not), then run it with hai project start.

The session token is stored in ~/.hedwigai/config.json and is the Bearer for the hedwigai sandbox API. For CI/headless use, set HEDWIGAI_AUTH_TOKEN and skip signin. Requires Node 18+.

SDK

Installation

npm install hedwigai
# or
yarn add hedwigai

Getting Started

1. Import hedwigai into your TypeScript project:

   import { Library } from "hedwigai";

2. Create a library instance for accessing your private knowledge graph:

   const library = new Library();
   library.signIn(username, password).then(() => {
    console.log("🙋 Sign In Successful!")
   })

3. Create data pipelines with a single line of code

• Loading files into the library:

  const identifier = library.indexFile(file);

• Search for the file you want using a text prompt

  const prompt = "catalog bedroom furniture items for a user who recently visited Bali";
  const displayItems = library.findImage(prompt);

  const prompt = "seek moment where virat kohli hit the winning six";
  const displayItems = library.findVideo(prompt);

• Generate the file you want using a text prompt and a target scope

  const identifier = "generate a new image using the furniture items for reference. Replace floral accents and replace with geometric patterns";
  const displayItems = library.generate(prompt, displayItems);

  It returns the file references along with auto-annotated relationships to all other entities in the connected datalake.

Live Support & Assistance

For live support, reach out to us on discord or talk to our AI assistant

API Reference

For detailed documentation and examples, refer to our API Reference.

Licensing Information

• hedwigai is available for commercial use under a custom commercial license.
• Businesses are required to obtain a valid commercial license to use this package in commercial applications.

Pricing Structure

• We offer a flexible pricing structure for businesses. You can reach out to us on [email protected] for information.

Start building with hedwigai today! Sign up at https://www.hedwigai.com/

Happy coding! 🚀✨