# inference.sh

> An agent runtime platform that handles durable execution, tool orchestration, observability, and infrastructure so developers can run reliable AI agents in production.

inference.sh is an agent runtime platform that eliminates the infrastructure burden of running AI agents in production. It provides durable execution, 150+ pre-built tool integrations, real-time observability, and human-in-the-loop controls so developers can focus on what their agents do rather than how to keep them running. The platform supports no-code, low-code, and full API workflows, making it accessible to builders at every level. It is built around a trust-first philosophy where every action is traceable, failures are graceful, and automation is never a black box.

- **Durable Execution**: *Event-driven, checkpoint-based execution ensures agents resume from the last successful step after failures, timeouts, or restarts — no lost state.*
- **Tool Orchestration**: *Access 150+ apps as agent tools via a single API, with structured execution, approval gates, and full visibility into what ran.*
- **Observability**: *Every tool call, decision, and action is automatically traced and streamed in real time — no instrumentation required.*
- **Human-in-the-Loop**: *Add approval gates with a single flag; agents pause, show their intended action, and wait for confirmation before proceeding.*
- **Deep Agents (Sub-Agents)**: *Orchestrator agents can spawn specialist sub-agents as tools, delegating tasks and collecting structured results back up the chain.*
- **Dynamic Widgets**: *Agents generate interactive UI elements — forms, charts, selections — rendered inline in the chat interface.*
- **Pay-Per-Execution Pricing**: *Credits-based model with no idle costs; tiers unlock automatically based on cumulative usage.*
- **Custom App Creation**: *Scaffold, code, and deploy your own apps using the CLI and Python or JavaScript SDKs; schemas automatically become tool parameters.*
- **Visual Workflow Builder**: *Drag-and-drop flow editor chains apps into multi-step pipelines, deployable as a single callable app.*
- **Real OAuth Integrations**: *Durable, encrypted integrations with Google, Slack, Discord, X.com, Microsoft, Salesforce, Notion, and more — with automatic token refresh.*
- **Bring Your Own Keys (BYOK)**: *Use your own GCP, Azure, or AWS billing and credits for AI models.*
- **Agentic Payments (x402)**: *Managed wallets and budget controls let agents make programmatic payments autonomously via the x402 protocol.*
- **Self-Hosted Option**: *Deploy inference.sh in your own VPC or on-premises for maximum data control and privacy.*
- **Python & JavaScript SDKs**: *Install via `pip install inferencesh` or `npm i @inferencesh/sdk` to create, manage, and run agents fully programmatically.*

## Features
- Durable execution with checkpoint-based state persistence
- 150+ pre-built tool integrations via single API
- Real-time observability and automatic tracing
- Human-in-the-loop approval gates
- Deep agents / sub-agent orchestration
- Dynamic inline UI widgets
- Visual drag-and-drop workflow builder
- Custom app creation with CLI
- Python and JavaScript SDKs
- Real OAuth integrations with token refresh
- Bring Your Own Keys (BYOK)
- Agentic payments via x402 protocol
- Self-hosted / on-premises deployment
- Pay-per-execution credits model
- Webhooks and async callbacks
- Built-in key-value memory per conversation
- Multi-step planning with interruption resume
- Structured output for orchestrators

## Integrations
Google (OAuth, Service Account, GCP, Vertex AI, BigQuery, Cloud Storage), Slack, Discord, X.com (Twitter), Microsoft (Azure), Salesforce, Notion, AWS, OpenAI, Anthropic, Google Gemini, Meta (Llama), Mistral, DeepSeek

## Platforms
IOS, WEB, API, DEVELOPER_SDK

## Pricing
Open Source, Free tier available

## Links
- Website: https://inference.sh
- Documentation: https://inference.sh/docs
- Repository: https://github.com/inference-sh
- EveryDev.ai: https://www.everydev.ai/tools/inference-sh