Skip to content

HaruHunab1320/agentic

Repository files navigation

Agentic

Agentic is an experimental multi-agent development environment that lets you orchestrate Claude, Gemini, Aider, and now a preview Codex research agent from a single CLI. The project focuses on fast iteration, rich observability, and composable workflows that stitch together analysis, implementation, testing, and long-running research loops.

Highlights

  • CLI orchestratoragentic bootstraps shared memory, task routing, and intelligent coordination (src/agentic/core).
  • Rich documentation – Phase plans, outlines, and architecture notes live in docs/ while historical analyses are archived under docs/archive/.
  • Curated examples – Legacy frontend/backend demos now reside in examples/ so the Python package stays clean.
  • Codex research agent (preview) – A new agent type scaffolds unattended research sessions and paves the way for automated experiments.

Getting Started

  1. Create a virtual environment and install the package in editable mode:
    python -m venv .venv
    source .venv/bin/activate
    pip install -e .
  2. Run the CLI:
    agentic --help
  3. Explore the documentation starting with docs/ARCHITECTURE.md and the phase briefs in docs/Phase*.md.

Repository Layout

docs/                    → design notes, outlines, and archived reports
examples/                → sample projects (full-stack, vite, snippets, archives)
src/agentic/             → Python package with orchestrator, agents, and models
tests/                   → pytest suite covering agents, coordination, and QA

Codex Research Agent (Preview)

  • AgentType.CODEX_RESEARCH is registered with a placeholder CodexResearchAgent that maps out the control loop for future Codex integrations.
  • The agent records preliminary notes, emits a research_finding discovery, and returns a summary output while running in preview mode (no external API calls yet).
  • See docs/outlines/RESEARCH_AGENT_PLAN.md for the roadmap to full automation, including sandboxed experiment runners and outcome verification.

Spawn it manually once the orchestrator is configured:

agentic spawn research --model codex-latest

(API wiring is still pending; the command reports planned work.)

Documentation Guide

  • Outlines & Roadmapsdocs/outlines/
  • Phase Execution Plansdocs/Phase1.mddocs/Phase6.md
  • Architecture & Specsdocs/ARCHITECTURE.md, docs/verification-loop.md
  • Historical Analysesdocs/archive/analysis/

Contributing

The repo is intentionally exploratory. File issues or drop notes in docs/outlines/ if you identify new workflows or want to help implement the Codex research pipeline.

About

Multi-agent AI development workflows from a single CLI

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published