Fresh workers each cycle
Sends a bounded goal to fresh Codex workers so each cycle starts with clean context instead of continuing from a long, tangled chat thread.
AgentLoop
ReclamarAgentLoop is a local coding loop tool for unattended work in a local repository, with fresh workers, an independent critic, and project standards.
Overview
AgentLoop is a local coding loop tool that automates handoffs between planning, implementation, and review. The homepage describes it as “unattended local coding loops” with fresh workers, an independent critic, and standards applied on every cycle.
The intended workflow is to set the goal once, keep the rubric in GUIDELINES.md, and let the system keep iterating on the local project until the critic and your standards agree the work is done. It is presented as a local-first engine that runs in the repository where the code lives.
Features
Independent critic
Uses a separate critic process to review results independently, so passing tests are treated as evidence rather than the only stopping condition.
Files as memory
Carries project files, logs, and feedback forward between cycles so the work can continue in the local repository without the user relaying every message.
Rubric in GUIDELINES.md
Loads the definition of done from GUIDELINES.md, letting the project’s standards persist across iterations instead of living only in prompts.
Local-first setup
Runs locally with Node.js 18+, Git, and an authenticated Codex CLI, and the source says it does not require an npm install or hosted workspace.
Cycle observability
Provides observability through cycle logs, verdicts, transcripts, cost, file handoffs, and an interactive replay/dashboard view.
Use cases
Feature delivery
Assign a scoped feature and let the loop build, review, and refine it against acceptance criteria stored in your project files.
Migrations and refactors
Use the worker/critic cycle for code changes that must preserve existing behavior while updating data handling, routes, or file formats.
Hardening work
Run an unattended pass on brittle areas of the codebase to add fixes, regression coverage, and documentation alignment.
Product polish
Hand off a UI, copy, or developer-experience cleanup task when you want iterative polishing without staying in the loop for each step.
Pros and Cons
Pros
- Automates the repeated handoff between planner, worker, and critic in a local repository.
- Keeps project standards in a persistent file rather than requiring prompt re-entry every cycle.
- Runs with fresh worker context and a separate critic process, which the site presents as a way to catch issues tests can miss.
- Supports unattended iteration while still exposing logs, verdicts, transcripts, and handoff details for review.
- Does not require a hosted workspace according to the homepage copy.
Cons
- The source does not include a working pricing page, so cost and plan details are not confirmed.
- The coverage available here is limited to homepage and pricing-page text, so integrations, supported outputs, and team workflows are not documented in the provided material.
- The product depends on Codex CLI and local prerequisites, which may narrow who can run it immediately.
FAQ
What does AgentLoop do?
AgentLoop is described as running unattended local coding loops: you set a goal and standards once, then it sends work to fresh Codex workers, has an independent critic review the result, and carries evidence forward into the next cycle.
What do I need to run it locally?
The source says it runs where your code lives and requires Node.js 18+, Git, and an authenticated Codex CLI. It also says there is no npm install and no hosted workspace.
How does the workflow work?
The homepage describes a workflow where you define the finish line in ChatGPT, put your rubric in GUIDELINES.md, and let AgentLoop handle the worker/critic loop until the work passes your standards.
What kind of work is it for?
The site positions it for bounded coding work such as feature work, migrations, hardening passes, and product polish, especially when you want to avoid manually babysitting each handoff.
What does AgentLoop cost?
The pricing URL in the provided source returns a 404, so the collected pages do not confirm any pricing details.
Quick Facts
- Category
- Developer Tool
- Primary workflow
- Unattended local coding loops with worker and critic cycles
- Platform
- Local repository / self-hosted execution
- Requirements
- Node.js 18+, Git, authenticated Codex CLI
- Source domain
- agentloop.tools
- Pricing
- Not confirmed; pricing page returned 404 in the collected source
Alternativas a AgentLoop
Orca
Orca is an Agent Development Environment for shipping with coding agents, running multiple CLI agents in parallel across isolated worktrees, with desktop and mobile workflows.
Firebase Studio
Firebase Studio is a web-based workspace for full-stack app development with Gemini-assisted coding, app previews, cloud emulators, collaboration, and browser deployment.
AI Magicx
AI Magicx is a unified AI workspace for chat, image, video, voice, music, email and developer tasks, helping teams and creators manage multiple models in one place.
Paper
Paper is a design tool that connects canvas, code, and AI agents so teams can create, share, and ship work in one workflow. Includes desktop app and MCP access.
blop
blop is a QA agent that writes browser tests as code in your repo, runs them in CI, clusters repeated failures, and can open PRs to fix broken tests.
RLAMA
RLAMA is a local AI platform for building RAG systems and intelligent agents on macOS, Linux, and Windows, with HTTP API support.