Claude Code is a great tool. But it’s tied to Anthropic’s API, proprietary, and can’t be embedded into your own product. What if you could build the same experience — with any AI provider, open source, and fully under your control?

With Robota SDK, you can. Here’s how.

The Fastest Path: Install and Run

npm install -g @robota-sdk/agent-cli
robota

That’s it. On first run, you’ll be asked to choose a provider (Anthropic, OpenAI, DeepSeek, Gemini, or a local LM Studio model) and enter your API key. Then you get the same interactive coding assistant experience — file reading, editing, bash execution, multi-turn conversations — all in your terminal.

If you already use Claude Code, the CLI can read your project’s .claude/settings.json and CLAUDE.md after you trust the workspace. In Restricted mode it does not load those project files.

Build a Custom Agent in 50 Lines

The CLI is built on top of a programmable SDK. Here’s a minimal coding assistant you can embed in your own tooling:

import { InteractiveSession } from '@robota-sdk/agent-framework';
import { createNodeFrameworkHost } from '@robota-sdk/agent-framework/node';
import { createDefaultTools } from '@robota-sdk/agent-tools/defaults';
import { AnthropicProvider } from '@robota-sdk/agent-provider/anthropic';
 
const provider = new AnthropicProvider({
  apiKey: process.env.ANTHROPIC_API_KEY,
});
 
const session = new InteractiveSession({
  host: createNodeFrameworkHost({ createDefaultTools }),
  cwd: process.cwd(), // project context
  provider,
  permissionMode: 'default',
});
 
// Stream output
session.on('text_delta', (delta) => process.stdout.write(delta));
session.on('complete', () => console.log('\n'));
 
// Submit a task
await session.submit('List all TODO comments in this project and suggest fixes');

This session uses tools inside cwd, applies the default permission mode, and maintains multi-turn conversation history. Without a projectAccess trust decision, it runs Restricted and does not load project instructions, settings, or memory.

Switch to a Cheaper Provider in One Line

This is where Robota’s multi-provider design really shines. Anthropic raised prices? Switch to DeepSeek — same code:

import { InteractiveSession } from '@robota-sdk/agent-framework';
import { createNodeFrameworkHost } from '@robota-sdk/agent-framework/node';
import { createDefaultTools } from '@robota-sdk/agent-tools/defaults';
import { AnthropicProvider } from '@robota-sdk/agent-provider/anthropic';
// Before: Anthropic claude-sonnet-4-6
// const provider = new AnthropicProvider({ apiKey: process.env.ANTHROPIC_API_KEY });
 
// After: DeepSeek (10-30x cheaper for many tasks)
import { DeepSeekProvider } from '@robota-sdk/agent-provider/deepseek';
const provider = new DeepSeekProvider({ apiKey: process.env.DEEPSEEK_API_KEY });
 
// The rest of your code is unchanged
const session = new InteractiveSession({ cwd: process.cwd(), host: createNodeFrameworkHost({ createDefaultTools }), provider });

No rewrites. No migration. The provider interface is identical across all 8 supported providers.

Embed in Your Own Product

The real power is in the transport layer. The same InteractiveSession can be exposed over:

  • HTTP/REST — for server-side integrations
  • WebSocket — for real-time browser UIs
  • MCP — to be called by Claude Code or other MCP-compatible tools
  • Headless — for CI/CD and scripted use cases

Install @robota-sdk/agent-transport, hono and @hono/node-server to serve HTTP requests.

import { createHttpTransport } from '@robota-sdk/agent-transport/http';
import { InteractiveSession } from '@robota-sdk/agent-framework';
import { createNodeFrameworkHost } from '@robota-sdk/agent-framework/node';
import { createDefaultTools } from '@robota-sdk/agent-tools/defaults';
 
import { AnthropicProvider } from '@robota-sdk/agent-provider/anthropic';
import { serve } from '@hono/node-server';
 
const provider = new AnthropicProvider({ apiKey: process.env.ANTHROPIC_API_KEY });
const session = new InteractiveSession({ cwd: process.cwd(), host: createNodeFrameworkHost({ createDefaultTools }), provider });
const transport = createHttpTransport();
transport.attach(session);
await transport.start();
await session.whenInitialized();
console.log('token:', transport.getAdmissionToken());
serve({ fetch: transport.getApp().fetch, port: 3000 });
// Your AI coding assistant is now an API

Try Without an API Key

If you want to experiment without spending money, use LM Studio:

  1. Install LM Studio
  2. Download any model (Llama, Mistral, Phi, etc.)
  3. Start the local server in LM Studio
  4. Run robota — select “LM Studio” when prompted

No API key. No cost. Same experience.

What’s Next