Choose CLI, SDK, MCP, or HTTP

The web app, MCP, the HTTP API, the SDKs and the CLI reach the same workspace, credits and Library. Pick the one that matches how you work.

Updated 2026-10-22

Pick the way in that fits your work

Wavemaker has one service and several ways to reach it. Most people only need the web app.

If you aren't sure, start in the web app. Everything you make there is available to the other ways in later.

What every way in shares

All of them act on the same workspace:

• Credits. Work started from any of them spends from the same workspace balance. See Credits.
• Starter credits are the exception. The free credits a new account starts with can't be spent through MCP or a developer credential until the workspace has paid for credits or a plan. Use them in the web app.
• Library. Files you upload and media Wavemaker makes are owned by the workspace, wherever they came from. See Library.
• Conversations. The web chat and the API's hosted-agent sessions are the same thing underneath. A session your software starts without naming its own folder becomes a conversation you can open in the web app.
• Rules. The same content policies, spending limits and permissions apply everywhere. See Content policies.

A credential belongs to one workspace. To work in another workspace, use a credential made there.

The web app

The web app at /app is where you write a brief, add references, choose quality and models, follow progress, answer the agent's questions and review results. It is the only way in that shows previews, side-by-side comparisons, history and sharing controls. See Your first creative conversation.

Pick the web app when a person is directing the work and reviewing what comes back.

MCP: Wavemaker inside your AI assistant

MCP (Model Context Protocol) lets an AI assistant call Wavemaker's tools while you chat with it. You keep working in the assistant you already use, and it can search, generate, edit, review and deliver media in your workspace.

1. In your assistant's connector or MCP settings, add a custom server at https://api.wavemaker.io/mcp.
2. When the assistant opens the Wavemaker sign-in page, sign in and approve access.
3. Ask the assistant for something that uses Wavemaker.

Leave client ID and secret empty; the sign-in handles them. Setup steps for Claude, ChatGPT, Cursor and other hosts are in the MCP guide.

Pick MCP when you want an assistant you already use to make or edit media, or to fold media work into a coding or writing task.

To see or remove an assistant's access, open Settings › Connections and choose the App grants tab. Revoke grant stops future requests; requests already running finish.

HTTP API: Wavemaker inside your software

The HTTP API is for developers. Your software sends requests to https://api.wavemaker.io with a developer credential. It can work two ways:

• Hosted agent sessions. Send a brief, as you would in the web app, and the Wavemaker agent plans and makes the work. Your software follows progress and collects the finished files.
• Tools. Call the same twelve tools the agent uses, one at a time: search, read, write, generate, edit, compose, watch, listen, compare, review, deliver and run. Your software makes the decisions.

You can ask what a request will cost before running it, and a request repeated with the same request ID returns the original result instead of charging twice. See the HTTP API reference.

Pick the API when Wavemaker is part of a product or an automated workflow, such as making a batch of ads from a feed, with no one clicking through each one.

SDK

The SDKs wrap the HTTP API for Python, TypeScript and Go, adding uploads, verified downloads and waiting for jobs: wavemaker-sdk for Python, @wavemakerdotio/sdk for Node.js and the Go module github.com/AdwaveInc/wavemaker-go. Official packages appear after a public registry release, and SDKs shows whether that has happened. Until then, use the HTTP API directly.

CLI

The Wavemaker CLI runs from your terminal. It signs in through your browser and works with the same workspace you use on the web. Until its first release, the Download page shows Not released yet and lists the planned builds for macOS, Linux and Windows. See the CLI guide for how it signs in and which commands it offers.

The wm client inside run

Code you run with the run tool already has the wm command and Python client, with a token for the project: wm search "image to video", wm generate --json '{...}' --request-id {key}. It exists only inside that sandbox and is not something to install; outside it, use the CLI, an SDK, HTTP or MCP.

Create a developer credential

The HTTP API needs a developer credential. MCP sign-in doesn't.

1. Open Settings › Connections and choose the Developer setup tab.
2. Under Developer credentials, select Create credential.
3. Enter a Name, such as My local scripts, and choose Expires in: 1 day, 7 days or 30 days.
4. Choose what it may do. Read projects and results is always included. Add Create and edit source to upload and change files, and Run work to start anything that spends credits.
5. Create it, then copy the credential from Save your credential.

A credential can only have permissions your own account has. To create one you need edit access to the workspace.

Replace or revoke a credential

The Developer credentials table lists each credential's Name, Access, Expires and Status.

• Replace makes a new credential. The old one stops working when the replacement is created.
• Revoke blocks future requests made with it. Work already in progress is handled separately.

Changing your password revokes your developer credentials and app grants. See Credential guide.

Which one should I use?

Developer overview · MCP guide · HTTP API

Related guidance

• HTTP API
• Client credentials and provider keys
• Connect an external agent with MCP
• Connect to Wavemaker
• Python, TypeScript and Go clients
• Your first creative conversation