rivet-dev/agents

A durable process per agent, with memory that survives restarts

What it solves

This project provides the infrastructure needed to run and manage AI agents as durable actors, as well as providing secure environments (sandboxes) for coding agents to execute tasks.

How it works

It consists of two primary components: Rivet Agents, which allows agents like Pi to be run as durable Rivet Actors, and a Sandbox Agent that enables the control of coding agents within isolated sandboxes via HTTP.

Who it’s for

Developers building AI agents who need a way to deploy them as durable actors or provide them with a secure, controllable execution environment for code.

Highlights

  • Durable actor-based agent infrastructure
  • Support for running specific agents like Pi
  • Sandbox environments for coding agents with HTTP control

Related

  • Project
    caezium/burrow

    Burrow is a free, open‑source macOS (stable) and Windows (preview) utility that bundles a Rust engine to provide junk‑file cleanup, dev‑artifact purging, installer sweeping, app uninstall, disk‑usage treemap, live system monitoring, and long‑term metric history. It also runs an MCP JSON‑RPC/HTTP server so AI agents (Claude Code, Cursor, etc.) can query the machine’s state and, with user consent, invoke these maintenance actions.

  • Project
    Flux159/mcp-server-kubernetes

    MCP Server Kubernetes is a Node‑based service that exposes a rich set of kubectl and Helm operations as MCP tools, letting AI assistants (Claude, Codex, VS Code, etc.) manage a Kubernetes cluster safely (including a read‑only mode) and optionally emit OpenTelemetry traces.

  • Project
    BAI-LAB/MemoryOS

    MemoryOS is a memory management system for AI agents that uses a hierarchical storage architecture to maintain short, mid, and long-term personalized memory for more coherent interactions.

  • Project
    archestra-ai/OpenAPPA

    OpenAPPA is a deterministic, policy‑driven guardrail that sits between LLM agents and their tools, checking each tool call against a TOML‑based permissions policy to prevent unauthorized data flows. It offers a CLI for policy validation, integrations (e.g., Claude Code, Archestra proxy), and shows strong security benchmarks while maintaining high task‑completion rates.