diff --git a/readme.md b/readme.md
index b923f8a..aa3f7c1 100644
--- a/readme.md
+++ b/readme.md
@@ -49,6 +49,19 @@ client.createCompletion(params: CompletionParams): Promise<{
 }>
 ```
 
+To get a streaming response, use the `streamCompletion` method.
+
+```ts
+client.streamCompletion(params: CompletionParams): Promise<
+    ReadableStream<{
+      /** The completion string. */
+      completion: string;
+      /** The raw response from the API. */
+      response: CompletionResponse;
+    }>
+  >
+```
+
 ### Create Chat Completion
 
 See: [OpenAI docs](https://beta.openai.com/docs/api-reference/chat) | [Type definitions](/src/schemas/chat-completion.ts)
diff --git a/src/fetch-api.ts b/src/fetch-api.ts
index 086df4c..c1344a4 100644
--- a/src/fetch-api.ts
+++ b/src/fetch-api.ts
@@ -4,7 +4,7 @@ import { OpenAIApiError } from './errors';
 
 const DEFAULT_BASE_URL = 'https://api.openai.com/v1';
 
-export interface FetchOptions extends Options {
+export interface FetchOptions extends Omit<Options, 'credentials'> {
   credentials?: string;
 }
 
diff --git a/src/openai-client.ts b/src/openai-client.ts
index 13dc964..4f45e69 100644
--- a/src/openai-client.ts
+++ b/src/openai-client.ts
@@ -12,10 +12,10 @@ import type { FetchOptions } from './fetch-api';
 import type {
   ChatCompletionParams,
   ChatCompletionResponse,
-  ChatResponseMessage} from './schemas/chat-completion';
-import {
-  ChatCompletionParamsSchema
+  ChatResponseMessage,
 } from './schemas/chat-completion';
+import { ChatCompletionParamsSchema } from './schemas/chat-completion';
+import { StreamCompletionChunker } from './streaming';
 
 export type ConfigOpts = {
   /**
@@ -91,6 +91,49 @@ export class OpenAIClient {
     return { completion, response };
   }
 
+  /**
+   * Create a completion for a single prompt string and stream back partial progress.
+   * @param params typipcal standard OpenAI completion parameters
+   * @returns A stream of completion chunks.
+   *
+   * @example
+   *
+   * ```ts
+   * const client = new OpenAIClient(process.env.OPENAI_API_KEY);
+   * const stream = await client.streamCompletion({
+   *   model: "text-davinci-003",
+   *   prompt: "Give me some lyrics, make it up.",
+   *   max_tokens: 256,
+   *   temperature: 0,
+   * });
+   *
+   * for await (const chunk of stream) {
+   *   process.stdout.write(chunk.completion);
+   * }
+   * ```
+   */
+  async streamCompletion(params: CompletionParams): Promise<
+    ReadableStream<{
+      /** The completion string. */
+      completion: string;
+      /** The raw response from the API. */
+      response: CompletionResponse;
+    }>
+  > {
+    const reqBody = CompletionParamsSchema.parse(params);
+    const response = await this.api.post('completions', {
+      json: { ...reqBody, stream: true },
+      onDownloadProgress: () => {}, // trick ky to return ReadableStream.
+    });
+    const stream = response.body as ReadableStream;
+    return stream.pipeThrough(
+      new StreamCompletionChunker((response: CompletionResponse) => {
+        const completion = response.choices[0].text || '';
+        return { completion, response };
+      })
+    );
+  }
+
   /**
    * Create a completion for a chat message.
    */
@@ -111,6 +154,31 @@ export class OpenAIClient {
     return { message, response };
   }
 
+  async streamChatCompletion(params: ChatCompletionParams): Promise<
+    ReadableStream<{
+      /** The completion message. */
+      message: ChatResponseMessage;
+      /** The raw response from the API. */
+      response: ChatCompletionResponse;
+    }>
+  > {
+    const reqBody = ChatCompletionParamsSchema.parse(params);
+    const response = await this.api.post('chat/completions', {
+      json: { ...reqBody, stream: true },
+      onDownloadProgress: () => {}, // trick ky to return ReadableStream.
+    });
+    const stream = response.body as ReadableStream;
+    return stream.pipeThrough(
+      new StreamCompletionChunker((response: ChatCompletionResponse) => {
+        const message = response.choices[0].delta || {
+          role: 'assistant',
+          content: '',
+        };
+        return { message, response };
+      })
+    );
+  }
+
   /**
    * Create an edit for a single input string.
    */
diff --git a/src/schemas/chat-completion.ts b/src/schemas/chat-completion.ts
index 43517c1..dbae332 100644
--- a/src/schemas/chat-completion.ts
+++ b/src/schemas/chat-completion.ts
@@ -76,4 +76,6 @@ export type ChatCompletionResponseChoices = {
   index?: number;
   finish_reason?: string;
   message?: ChatResponseMessage;
+  /** Used instead of `message` when streaming */
+  delta?: ChatResponseMessage;
 }[];
diff --git a/src/streaming.ts b/src/streaming.ts
new file mode 100644
index 0000000..7eaf023
--- /dev/null
+++ b/src/streaming.ts
@@ -0,0 +1,71 @@
+/** A function that converts from raw Completion response from OpenAI
+ * into a nicer object which includes the first choice in response from OpenAI.
+ */
+type ResponseFactory<Raw, Nice> = (response: Raw) => Nice;
+
+/**
+ * A parser for the streaming responses from the OpenAI API.
+ *
+ * Conveniently shaped like an argument for WritableStream constructor.
+ */
+class OpenAIStreamParser<Raw, Nice> {
+  private responseFactory: ResponseFactory<Raw, Nice>;
+  onchunk?: (chunk: Nice) => void;
+  onend?: () => void;
+
+  constructor(responseFactory: ResponseFactory<Raw, Nice>) {
+    this.responseFactory = responseFactory;
+  }
+
+  /**
+   * Takes the ReadableStream chunks, produced by `fetch` and turns them into
+   * `CompletionResponse` objects.
+   * @param chunk The chunk of data from the stream.
+   */
+  write(chunk: Uint8Array): void {
+    const decoder = new TextDecoder();
+    const s = decoder.decode(chunk);
+    s.split('\n')
+      .map((line) => line.trim())
+      .filter((line) => line.length > 0)
+      .forEach((line) => {
+        const pos = line.indexOf(':');
+        const name = line.substring(0, pos);
+        if (name !== 'data') return;
+        const content = line.substring(pos + 1).trim();
+        if (content.length == 0) return;
+        if (content === '[DONE]') {
+          this.onend?.();
+          return;
+        }
+        try {
+          const parsed = JSON.parse(content);
+          this.onchunk?.(this.responseFactory(parsed));
+        } catch (e) {
+          console.error('Failed parsing streamed JSON chunk', e);
+        }
+      });
+  }
+}
+
+/**
+ * A transform stream that takes the streaming responses from the OpenAI API
+ * and turns them into useful response objects.
+ */
+export class StreamCompletionChunker<Raw, Nice>
+  implements TransformStream<Uint8Array, Nice>
+{
+  writable: WritableStream<Uint8Array>;
+  readable: ReadableStream<Nice>;
+
+  constructor(responseFactory: ResponseFactory<Raw, Nice>) {
+    const parser = new OpenAIStreamParser(responseFactory);
+    this.writable = new WritableStream(parser);
+    this.readable = new ReadableStream({
+      start(controller) {
+        parser.onchunk = (chunk: Nice) => controller.enqueue(chunk);
+        parser.onend = () => controller.close();
+      },
+    });
+  }
+}
diff --git a/tsconfig.json b/tsconfig.json
index 37d6397..040a9bc 100644
--- a/tsconfig.json
+++ b/tsconfig.json
@@ -10,7 +10,7 @@
     "esModuleInterop": true,
     "forceConsistentCasingInFileNames": true,
     "isolatedModules": true,
-    "lib": ["es2021"],
+    "lib": ["es2021", "DOM"],
     "module": "commonjs",
     "moduleResolution": "node",
     "outDir": "dist",