persistent-agents

OpenRouter for agents and sandbox environments. Provides a control plane and configuration options for persistent agent workers. Route runtime + substrates, provision specialist presets, and supervise your fleet.

OpenClaw-oriented workspace template for AI-agent continuity: file-based memory, self-identity docs, and operating conventions.

Open-source security-first runtime for AI agents that learn and act across sessions.

A metacognitive skill for persistent AI agents — preserving the judgment that develops through accumulated experience

Judgment preservation in persistent AI agents. Toroidal routing engine + Emergent Judgment protocol. 17/17 tests passing. 1,563× improvement in context-sensitive routing validated through 11 experiments.

A portable set of Markdown role definitions for lean founder-led companies using persistent AI agents.

TwinBench: Open benchmark for personal AI assistant runtimes.

Research package for Agent Concretization and Schema Sandbox.

Local multi-cycle agent runtime: supervised ticks, contracted cycle runner, watchdog recovery, rolling segments, and restart-safe on-disk continuity. Public control plane for long-horizon local AI.

Theory-first architecture for agents centered on continuous existence rather than task completion.

Agent Operating System running locally in the browser: a persistent, tool-driven AI that reasons in an agentic loop, executes skills, manages memory/state, and evolves over time with optional sync and strict governance.

Governed operating loops and portable capability Capsules for persistent AI agents

Coven-native familiar runtime and orchestration protocol for persistent AI identities.

Persistent Agent Pool MCP Server for Claude Code — Your Context Is Always Near.

Open-source memory and experience layer for persistent AI agents, with auditable decisions, evidence-based learning, and evolving identity.

A persistent agent that keeps thinking between conversations.

Durable single-host runtime for persistent agent teams: scheduling, causal recovery, delegation, and a local workbench for long-running Pi Harness sessions.

Wake past agent sessions as named, bounded participants—local-first, attributed, persistent, and MCP-native.