Deploy an Oasis Platform instance (Docker-based) or use the OasisHub hosted environment; the platform exposes a Django-backed REST API documented at oasislmf.github.io
Prepare exposure data in OED (Open Exposure Data) format — a standard CSV schema covering location, financial, and account-level data; validate the file against the OED schema before upload
POST the exposure file to the /analyses/portfolios/ upload endpoint; await the portfolio validation response confirming the data passes integrity checks
Create an analysis run by POSTing to the /analyses/ endpoint with the portfolio ID and target model ID; the model must be deployed as a Docker container registered with the platform
Start the analysis with a POST to /analyses/{id}/run/; poll GET /analyses/{id}/ until the status transitions to 'run_completed'
Retrieve the output files (GUL, IL, RI losses) via GET /analyses/{id}/output_file/; parse the CSV loss outputs and load into your aggregation or reporting pipeline
Known gotchas
Oasis is model-agnostic; you must separately obtain and deploy a compatible cat model (from the Oasis Model Library or a commercial provider) — the framework itself does not supply peril models
The OED standard is versioned; confirm that your exposure file version matches what the deployed model's data conversion components expect — a version mismatch causes silent data loss or job failure
Long-running analyses require the Celery worker queue to be healthy; monitor worker pod health separately from the API server, as API availability does not imply worker availability
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp