# Develop and extend AxonX

Keep research algorithms in plugins and reuse the framework's extension points for execution, records, and interfaces. Choose the layer you need to extend before changing code.

## Choose an extension path

| Goal                                           | Start here                                          | Contract to consult                                                             |
| ---------------------------------------------- | --------------------------------------------------- | ------------------------------------------------------------------------------- |
| Prepare a source checkout and run checks       | [Contributing](https://flowllm-ai.github.io/AxonX/en/development/contributing)            | Repository contribution rules                                                   |
| Implement a research Task                      | [Development and operations guide](https://flowllm-ai.github.io/AxonX/en/dev_guide) | [Task inputs, outputs, identity, and lifecycle](https://flowllm-ai.github.io/AxonX/en/reference/task-contracts) |
| Package and register a plugin                  | [Plugin management](https://flowllm-ai.github.io/AxonX/en/plugins/management)       | [Plugin manifest](https://flowllm-ai.github.io/AxonX/en/reference/plugin-manifest)                              |
| Extend components, Jobs, or asynchronous steps | [Framework extensions](https://flowllm-ai.github.io/AxonX/en/development/framework-extensions)     | [Architecture](https://flowllm-ai.github.io/AxonX/en/concepts/architecture)                                     |
| Build Studio features and artifact views       | [Studio development](https://flowllm-ai.github.io/AxonX/en/development/studio)                     | [Research artifacts](https://flowllm-ai.github.io/AxonX/en/reference/research-artifacts)                        |
| Integrate a service client                     | [Python reference](https://flowllm-ai.github.io/AxonX/en/reference/python)          | [HTTP API](https://flowllm-ai.github.io/AxonX/en/api/overview) and [events](https://flowllm-ai.github.io/AxonX/en/api/events)                   |

The development and operations guide is also bundled for optional built-in Agent loading. Its canonical `dev_guide.md` path remains available to external Skills and source links; focused pages own the deeper explanations and field references.

## Validate the research change

Discover the registered Task's schema, exercise a minimal execution in a temporary workspace, and inspect its output metadata and declared artifacts. Changes to feature timing, fields, or upstream requirements need compatibility checks before reusing earlier records.

Use [experiment design](https://flowllm-ai.github.io/AxonX/en/research/experiments) for algorithm comparisons and [external agents](https://flowllm-ai.github.io/AxonX/en/agent/external) for agent-driven development. The [Alpha158 Enhanced plugin](https://flowllm-ai.github.io/AxonX/en/plugins/alpha158-enhanced) demonstrates a separate implementation with feature-group switches and independent confirmation.

## Maintain the documentation

Update English and Chinese pages together. Keep algorithm details with plugins, general workflows in guides, and field definitions in references. The [documentation source guide](https://flowllm-ai.github.io/AxonX/en/docs) and [site maintenance guide](https://github.com/FlowLLM-AI/AxonX/blob/main/github-pages/README.md) explain navigation ownership, generation, and build checks.