Aileron is the runtime that lets your AI agent act in the real world. Send the email, post the Slack update, file the ticket. Your credentials stay in a local vault. Your approval gates the irreversible steps. Every action runs the same way every time, against the real API, inside a sandbox.
It complements Tool Calling and MCP rather than replacing them. The LLM still expresses intent through tool calls. MCP still exposes the catalog. Aileron is the layer that runs underneath, turning those calls into deterministic, auditable executions against your real systems.
Read the full story at docs.withaileron.ai
Everything lives at docs.withaileron.ai:
- Overview — what Aileron is and the pitch
- Getting Started — install, equip Gmail and Calendar actions, launch Claude Code through Aileron
- Concepts — the architecture layered: deterministic execution, actions, connectors, the vault, proof of control
- Architecture Decisions — the ADRs that ratify the post-Pivot architecture, including ADR-0012: Local Daemon Architecture
- API Reference — full OpenAPI specification
- Repository Layout — how the source tree is organized
- Building from Source — prerequisites and Taskfile entry points for contributors
- Running Tests — the test layers and how to run the by-hand black-box system-test suite
See LICENSE.