Skip to content

AxonX Documentation Map ​

AxonX is an agent-native harness for quantitative research, providing unified tools and a runtime for quantitative code development, experiment execution, and result analysis. You define the research question; the Agent implements it; the Harness manages execution and evidence.

The reference research plugins adapt Microsoft Qlib’s Alpha158 baseline, adding selectable factors and portfolio policies. AxonX also supports new research methods through plugins.

Choose your starting point ​

GoalStart hereOutcome
Use AxonX for the first timeQuickstart → StudioRun Demo and inspect its result, exit code, and record
Develop with an AgentAgent development and research promptPrepare source, tools, and an objective for plugin development
Run a research baselineResearch → Alpha158 workflowConnect data, training, prediction, and backtesting
Implement a research pluginDevelopment and operations guide → Plugin managementImplement Tasks, install and discover the plugin
Evaluate a changeExperiment design → Alpha158 caseCompare candidates, retain evidence, and identify limits
Manage execution servicesOperationsTrack Tasks, manage files and remote environments
Look up fields and protocolsReferenceFind interfaces, configuration, and extension contracts
Change AxonX itselfContributingExtend the framework or Studio and complete contribution checks

Read the project overview for positioning, plugins, and a case summary. The built-in demo needs no market data or model credentials. Research plugins need data; the built-in Agent needs service-side model configuration; external Agents use their host's model configuration.

Complete the demo first, then read about Jobs, Tasks and lifecycle, workspace, and lineage. See architecture for implementation principles.

The main research path is: connect an Agent → define an objective → develop a plugin → install in the execution service → run Tasks → inspect artifacts → compare and iterate. Agent development covers hosts, tools, and prompts; Research covers algorithm examples, plugin authoring, and evaluation; Contributing covers framework and Studio changes.

Each page has one navigation owner and cross-links to related steps. How-to guides explain procedures, Reference defines fields and contracts, plugin READMEs maintain algorithms, and case studies retain settings, metrics, reproduction commands, and Task provenance.

Reading conventions ​

English and Chinese pages use matching paths and share screenshots and diagrams. Replace example addresses, credentials, and Task IDs with actual values. Use the Job and Task schemas discovered on the connected service.

exec waits for the Task and returns business output plus a process exit code. Record its Task ID and metadata started_at for later observation. Lineage records do not automatically schedule a DAG. Reusing a name replaces a finished task directory; preserve distinct identities for experiment comparisons.

Start with the FAQ when something fails, then use troubleshooting and recovery to locate the failure in the service, Job, or Task layer.

Agent-native quant research.