# Talon

A cloud-native control plane that runs long-lived autonomous agents in production, with durable execution, declarative YAML configuration and a browser-based fleet view.

Talon provides infrastructure to operate long-lived AI agents in production. It keeps agent sessions durable across crashes, deployments and cold starts, while prompts, tools, workflows and policies stay explicit in a Kubernetes-style resource model of namespaces, templates, agents and sessions.

A browser-native view called Sightline lets operators inspect running sessions, resource state, schedules and execution history, and namespace-scoped MCP support lets agents use approved external tools and data sources. Agents, tools and policies are declared as YAML manifests and managed with a CLI.

Talon runs on cloud-agnostic primitives so it can be self-hosted, and the local stack is composed of a gateway, worker, Postgres and a Sightline UI orchestrated with Docker Compose; its CLI, server and worker binaries are written in Rust.

Website: https://impala.systems/talon/docs