long-running-agents

Persistent file-based planning for AI coding agents and long-running tasks. Crash-proof markdown plans, session recovery after /clear and compaction, per-turn re-injection against context rot, deterministic completion gate. Manus-style. Install from npm, the Claude Code plugin marketplace, or npx skills. Codex, Cursor, OpenCode, 60+ agents.

AutoScientists: Self-Organizing Agent Teams for Long-Running Scientific Experimentation

Give AI agents ambitious work without losing the plot. nac is an open-source harness for long-running tasks, using a central orchestrator, threads, and structured episodes to stay aligned with your intent.

Cayu is the runtime for long-horizon agents that need explicit environments, durable sessions, controlled tools, secrets boundaries, evals, and replay

Accountable long-running coding shifts for Cursor, Codex, and Claude Code. Safe autonomy, recovery, persistent state, and reviewable results a clear view of what’s done, in progress, parked, or blocked—without scrolling a marathon chat.

Run a company of AI agents on a server you own. Spin up named agents (claude, codex, pi…), put them on an org chart with a shared backlog, let them hand off work and ping your phone only when a human must decide. MIT.

A ralph harness for long-running AI coding tasks — a generator-evaluator loop that orchestrates Claude Code, OpenAI Codex & GitHub Copilot across repositories 🍩

Java-native runtime for long-running AI agents with live execution control, human approvals, and audit trails.

Local-first operating layer that decomposes long AI work into executable tasks, routes them across models and tools, and keeps human decisions in the loop.

Rate limit? Switching coding agents? Stele keeps write-ahead task state so work resumes in any harness with nothing to re-explain.

A framework for executing long-running AI agent tasks with structured feature lists and progress tracking.

Progress-aware execution for durable, grounded, token-efficient coding agents.

Pi extension that gives coding agents persistent /goal workflows with branch-aware state, source imports, progress tools, and explicit start handoffs.

An agent that hands the baton to itself: closes a milestone, writes its memory to disk, resumes with a clean context.

Long-running AI agents that adapt until outcomes converge.

Loop engineering for AI agents.

Embeddable, zero-dependency durable execution for agents and NHEs. Deterministic replay, retries, cycle detection, a token budget, and a multi-agent task board, with no sidecar service.

Build durable, governed, vendor-neutral agent harnesses with a visual Next.js studio, replayable runs, runtime approvals, and portable manifests.

MemoryLedger is a local-first memory transaction layer for long-running agents. It records not only what an agent remembers, but why that memory is believed, when and where it is valid, how it changed, and whether it is trustworthy enough to influence the next action.

watches the watchers: a monitoring system for long-running Claude Code agents