Inference Logoinference shell

Introduction

inference shell is an AI workspace where agents can actually do things. Put agents in charge of the busywork—they plan steps, pick the right tools or flows, and even generate small UI widgets when they need more input. You describe the goal. The agent handles the execution.

Key features:

  • 100+ pre-built apps: Access image generation, audio transcription, text analysis, and more from the Grid
  • Visual flow builder: Chain apps into multi-step workflows without code
  • AI agents with tools: Create assistants that can actually execute tasks, not just chat
  • Single API access: All apps accessible through one unified API with Python, JavaScript, and Swift SDKs
  • Real-time execution: Stream progress, logs, and results as they happen
  • Run anywhere: Cloud workers for convenience, private workers for data control
  • Model-native runtime: State, retries, monitoring—everything observable in one timeline

inference shell enables developers and teams to build AI-powered workflows without managing infrastructure. Create agents that use tools, build pipelines that process data, and deploy everything with a single command.

For web app features, see the Web app tour. For API access, see API & SDK.

Quick Example

Run an app with Python

python
1from inferencesh import inference23client = inference()  # reads INFERENCE_API_KEY, or pass api_key="inf_your_key"45result = client.run({6    "app": "infsh/stable-diffusion",7    "input": {"prompt": "A serene Japanese garden"}8})910print(result["output"]["images"][0]["uri"])

Run an app with JavaScript

typescript
1import { inference } from '@inferencesh/sdk';23const client = inference({ apiKey: 'inf_your_key' });45const result = await client.run({6  app: 'infsh/stable-diffusion',7  input: { prompt: 'A serene Japanese garden' }8});910console.log(result.output.images[0].uri);

Run an app from the CLI

bash
1belt login2belt app run infsh/stable-diffusion --input '{"prompt": "A serene Japanese garden"}'

On CI or a headless machine, set INFERENCE_API_KEY instead of running belt login. See CLI setup.

Create an agent (in the workspace)

  1. Go to Agents → New Agent
  2. Choose an LLM (GPT-4, Claude)
  3. Add tools (stable-diffusion, whisper, etc.)
  4. Start chatting—the agent uses tools automatically

Next Steps

nextwhat is inference shell?

we use cookies

we use cookies to ensure you get the best experience on our website. for more information on how we use cookies, please see our cookie policy.

by clicking "accept", you agree to our use of cookies.
learn more.