What problem it solves

Managing multiple coding agents across separate terminals leaves users responsible for scheduling, checking, and merging their work. Teams also need a practical way to route tasks across models, automate recurring work, and operate the same development process in interactive or headless environments.

Key features

• One workspace for every project — A single interface lists projects and conversations, keeps work running across tabs, and highlights questions, unread items, active tasks, and spending.

  Source: README

• Conversation-driven task factory — Describe one or several pieces of work in natural language. CodeAF can split them into tasks, run them in separate repository copies and branches, check completed work, and merge what passes.

  Source: README

• Task-specific subharnesses — Built-in specialists handle coding changes, security audits, and pull request reviews as tasks within the conversation. Users can also create and save specialized harnesses for recurring work.

  Source: README

• Per-task model crews — Each handed-off task can use worker, planner, and checker seats selected from connected providers. Users can inspect allowed models, set limits, pin a seat, or constrain all seats to open-weight models.

  Source: README

• Automations and headless execution — Reminders, scheduled work, and watches are configured conversationally and retain a history. The same workflow can run headlessly for CI, cron, scripts, and benchmark harnesses, returning JSON and an exit code.

  Source: README

Quick start

Copied verbatim from the project's README.

curl -fsSL https://agentfield.ai/get/codeaf | bash
codeaf

codeaf do "bump every dependency whose changelog is worth reading" --timeout 30m --json

Who it may suit

Inference: CodeAF may suit developers who want to coordinate multiple open-model coding tasks from one interface, and engineering teams that need recurring agent workflows in interactive or headless environments.

Topics to discuss

• How should coding agents divide work, verify results, and request human decisions?

• Which workflows are best suited to saved subharnesses or scheduled automations?

• How should teams balance model selection, spend limits, and task-specific routing?

Written by ReadmeX AI from the project's README, source and public GitHub data; it may contain mistakes.

Read the original README https://github.com/Agent-Field/CodeAF#readme

ReadmeX opened this community for a trending project; it is not run by the project. Maintainers can claim it from the community page.