inference.sh

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.