Vibe Coding

Vibe Coding means writing code by collaborating with an AI assistant — describing what you want in plain language and letting the AI generate, explain, or refactor the code for you. The Imbrace SDK ships an llms.txt file so any AI tool can instantly understand the SDK without hallucinating method names or argument shapes.

Setup

Before vibe coding, make sure the SDK is installed and your credentials are configured.

1. Install the SDK

TypeScript

npm install @imbrace/sdk

Python

pip install imbrace

2. Store your credentials

Create a .env file in your project root. The SDK does not auto-read environment variables — you pass them to the constructor in step 3.

IMBRACE_API_KEY=your_api_key_here
IMBRACE_ORGANIZATION_ID=your_org_id_here

See Authentication to learn when to use API Key vs Access Token, and Setup Guide for how to obtain an API key.

3. Initialize the client

TypeScript

import { ImbraceClient } from "@imbrace/sdk";

const client = new ImbraceClient({
  apiKey:         process.env.IMBRACE_API_KEY,
  organizationId: process.env.IMBRACE_ORGANIZATION_ID,
});

Python

import os
from imbrace import ImbraceClient

client = ImbraceClient(
    api_key=os.environ["IMBRACE_API_KEY"],
    organization_id=os.environ.get("IMBRACE_ORGANIZATION_ID"),
)

4. Grab llms.txt

Download or copy the file at https://developer.imbrace.co/llms.txt and drop it into your AI tool (see How to use it below).

---

What is llms.txt?

llms.txt is a plain-text file (similar to robots.txt) that gives AI models a compact, accurate summary of a library — its clients, resources, authentication, and common patterns. When you paste it into an AI context window, the model already knows the SDK and can write correct code on the first try.

File URL: https://developer.imbrace.co/llms.txt

How to use it

Claude (claude.ai or Claude Code)

1. Open a new conversation.
2. Paste the contents of llms.txt at the top of your message, then describe your task:

<context>
[paste llms.txt here]
</context>

Write a TypeScript snippet that streams a chat response from assistant "asst_abc"
and prints each text delta to the console.

Cursor / VS Code Copilot

Add the URL to your AI context via @ docs or the equivalent “add context” feature in your IDE. Cursor supports @URL directly:

@https://developer.imbrace.co/llms.txt

How do I upload a file and trigger embedding processing?

Any other LLM

Copy the raw file content and paste it at the start of your prompt before asking your question. Most LLMs with a 32k+ context window can ingest the full file without summarisation loss.

Example prompts

Once the AI has the llms.txt context, try prompts like:

• “Show me how to create an AI agent and stream a chat response in Python.”
• “Generate TypeScript code to list all embedding files and delete ones with status error.”
• “What’s the difference between streamChat and streamSubAgentChat?”
• “Write an Express.js auth proxy for the Chat Client, following the Integrations guide pattern.”

Tip

For long sessions keep llms.txt pinned as a system prompt or project instruction so it applies to every message automatically.

CLI for Coding Agents

The Imbrace CLI (@imbrace/cli) is also designed for coding agents. After installing the CLI, an AI assistant can run commands directly via the terminal without writing any SDK code:

# List boards
imbrace data-board list --json

# Create items
imbrace data-board create-item <boardId> --fields '<json>' --json

# List AI agents
imbrace ai-agent list --json

# Manage workflows
imbrace workflow list --json

# Document AI extraction agents
imbrace document-ai list --json

Resource commands (data-board, ai-agent, workflow, document-ai, orchestrator, guardrail) support --json for machine-readable output. Every command supports -h / --help for usage. See CLI Commands for the full reference.

Keep it up to date

The file is regenerated on every release. Re-fetch the URL if you upgrade the SDK to pick up new methods or changed signatures.