HLA-Compass documentation¶
Use HLA-Compass to discover governed scientific data, run available analysis modules and Nextflow pipelines, and inspect results. Start with the web application, or connect through the CLI, Python SDK, HTTP API or MCP.
You need an HLA-Compass account, organization membership and permission for the resources you use. Installing a package or connecting a chat app does not grant additional access. Use the environment supplied with your account; dev, staging and production have separate endpoints and resources.
Find the instructions you need¶
| Your task | Guide |
|---|---|
| Use the web application or make your first read-only request | Get started |
| Create an API key, choose the right authentication method or diagnose an access error | API and authentication |
| Connect Claude, ChatGPT, Mistral or a local MCP client | MCP and agent clients |
| Automate discovery and module execution with Python or the CLI | Python SDK |
| Build and publish a module | Build modules and publishing prerequisites |
| Register, configure and monitor a Nextflow pipeline | Build pipelines |
| Create catalog versions and ingest data | Data and ingestion |
| Follow a module integration example | Off Target Prediction over the API |
| Understand permissions, credentials and sharing | Security model |
Choose an interface¶
- Web application: explore Data, open Tools, inspect Runs and work in Canvas. Profile contains personal settings, API keys and connected chat apps.
- CLI and Python SDK: automate requests or develop and publish modules. Check the guide's SDK version and prerequisites before copying commands.
- HTTP API: integrate without Python. Use the target environment's
/v1/openapi.jsonand the authentication guide; bearer and API-key routes are distinct, and not every operation accepts every credential type. - MCP: let an external chat or agent client discover and use permitted HLA-Compass tools. This does not configure arbitrary external MCP servers inside HLA-Compass.
The interfaces enforce roles, organization boundaries and resource permissions. They expose different operations and authentication methods; consult the guide for the specific operation instead of substituting route prefixes or headers.
Core concepts¶
| Concept | Meaning |
|---|---|
| Organization | The tenant boundary for access, runs, storage and credits. |
| Catalog | A governed collection of scientific data. Visibility and allowed actions depend on its access policy. |
| Catalog version | A concrete data version. Use a ready version for a run and retain its identity for reproducibility. |
| Saved view | A selection or query definition. Check its source versions and whether it is pinned or follows the current version. |
| Module | A versioned analysis tool with declared inputs, outputs and an execution runtime. |
| Pipeline | A registered Nextflow workflow. Execution requires organization enablement, access and a supported compute profile. |
| Run | An execution record identified by a run ID. Use that exact ID to inspect status, results, provenance and credits. |
| Canvas notebook | A saved notebook artifact. Editing, execution and sharing are separate actions. |
Before executing work¶
Discover the actual resources available to your account rather than reusing IDs from another environment. Review inputs, source versions, compute choices and cost before confirming a module run. Retain the prepared plan and returned run ID. For a native pipeline, review its parameters and compute-profile charge before submission, as described in the pipeline guide.
A successful request is not evidence that a scientific analysis is valid. Check completion status, expected output content and recorded input provenance. The guides identify version-dependent features and verification limits; do not interpret an example as a readiness guarantee for every deployed module.