Introducing AgenticKode

We are excited to announce the public release of AgenticKode -- an open-source platform that automates software development tasks using AI agents, orchestrated through a structured pipeline that takes work from issue to pull request with minimal human intervention.

The Problem

AI coding assistants have become remarkably capable, but integrating them into real development workflows remains a manual process. Developers copy-paste context, babysit agent sessions, and manually create branches and PRs. The gap between "AI can write code" and "AI handles the full task lifecycle" is where AgenticKode steps in.

What AgenticKode Does

AgenticKode connects your issue tracker, git provider, and AI agents into a single automated pipeline. When a task arrives -- whether from GitHub Issues, Plane, GitLab, or a manual trigger -- AgenticKode:

1. Sets up a workspace on a remote server, cloning the repo and creating a feature branch
2. Analyzes the project to gather context about the codebase
3. Plans the implementation by decomposing the task into subtasks
4. Writes the code using your chosen AI agent (Claude CLI, Ollama, OpenHands, or custom agents)
5. Runs tests to validate the changes
6. Reviews the code with an AI reviewer for quality checks
7. Creates a PR and waits for human approval
8. Finalizes with notifications and cleanup

Every step is configurable. You can swap agents per step, inject custom instructions, define workflow templates from composable bash and agent primitives, and run multiple agents in parallel comparison mode. (When this post first shipped, AgenticKode used a fixed 8-phase pipeline; v0.5.0 replaced it with composable workflows — see Composable Agentic Workflows.)

Why Open Source

We believe AI-assisted development tooling should be transparent, extensible, and community-driven. By releasing under AGPLv3:

• You own your infrastructure. AgenticKode runs on your servers, with your keys, under your control. No data leaves your environment unless you choose to.
• You can extend it. The plugin architecture supports custom git providers, AI adapters, and notification channels. Build what your team needs.
• You can trust it. The full source is available for audit. Every workflow is documented and testable.

Getting Started

AgenticKode ships as a Docker Compose stack with a FastAPI backend and React frontend. Setup takes minutes:

git clone https://github.com/mechemsi/agentickode.git
cd agentickode
cp .env.example .env
docker compose up -d

Configure your git provider credentials, point it at a workspace server, and create your first run.

What Comes Next

This is just the beginning. Our roadmap includes expanded agent support, smarter planning with RAG-powered context retrieval, and tighter integrations with popular project management tools. We are building this in the open and welcome contributions from developers who share our vision of making AI-assisted development practical and reliable.

Check out the GitHub repository to get started, and join the Discussions to share your feedback.