OHI Studio

Bring Oracle Health Insurance into VS Code—for developers and AI agents.

OHI Studio gives developers and AI agents a controlled way to work with OHI environments, APIs, metadata, application artifacts, and runtime diagnostics. Use its visual tools directly or expose explicitly authorized capabilities through MCP.

One workspace for everyday OHI work

• MCP for AI agents — Let Codex and other compatible clients discover OHI capabilities, inspect authorized targets, call APIs, query logs, work with objects, and run guarded write or test flows.
• PostOHffice — Search APIs, compose requests, run repeatable scenarios, browse metadata, and inspect application logs from one persistent visual workbench.
• Dynamic Logic — Send, fetch, and test Groovy Dynamic Logic directly from the editor.
• Floorplans — Preview, send, and fetch Floorplans without leaving VS Code.
• OHI-aware navigation — Browse objects and APIs, generate a datamodel, and use hover help, signatures, highlighting, and autocomplete while editing Groovy.
• Multiple environments — Keep applications, authentication, active targets, and OHI landing pages organized in one place.
• Controlled access — Scope MCP access per environment and keep write-capable tools behind additional permissions.

MCP-powered OHI workflows

The local MCP server is OHI Studio's standout capability. It gives Codex and other compatible agents a structured way to understand and work with OHI instead of relying on copied payloads, browser context, or one-off scripts.

Agents can discover explicitly authorized environments and applications, inspect API metadata, search and call APIs, query application logs, browse OHI objects, download built Configuration Migration Tool (CMT) payloads, discover test values, and prepare test runs. CMT downloads are saved as new ZIP or XML files beneath the shared OHI Studio save folder and never overwrite an existing download. When write access is deliberately enabled, guarded tools can also update OHI data, send Dynamic Logic, generate datamodels, and run test workflows.

Access is controlled per environment with disabled, read-only, or read/write modes, plus per-tool permissions and a global control for write-capable tools. Stored credentials are used by the extension and are never returned to the agent.

For ohi_search_api_entity, agents must use an apiName and API family returned by that target's metadata. The MCP boundary rejects an invalid family/entity combination before issuing the entity search; it does not silently change a request. When the entity is advertised in the other family, the error recommends that family instead.

PostOHffice

PostOHffice is the visual companion to the MCP experience: a connected API workspace with API Search, API Client, API Runner, API Browser, and Application Logs. Its left navigation rail can collapse to a compact icon-only layout with hover descriptions.

Discover an entity, build a search, inspect the response, turn the request into a repeatable run, and follow the resulting logs—all against the target selected in VS Code.

Use Refresh API to reload both Generic APIs and Integration Points along with their cached definitions. API Browser keeps both catalogs in their correct modes, reloads a selected entity that remains available, and clears a selection that disappeared.

API Search keeps field metadata, generated previews, running status, and results scoped to the correct target and search tab. Cached entity fields and Browser definitions reopen immediately without another loading state; Refresh API deliberately invalidates them. Its completion menus support mouse selection or Arrow keys with Enter/Tab, and API Search and API Client support case-insensitive response Find from the response toolbar or with Ctrl/Cmd+F. PostOHffice text controls support Ctrl/Cmd+A, Ctrl/Cmd+Z, and the standard clipboard shortcuts consistently across embedded features. HTTP error status responses stay in the response pane with their status, headers, timing, and server body instead of being treated as extension failures.

Built around connected workflows

From code to OHI and back

Keep the most common Dynamic Logic and Floorplan actions beside the file you are editing. Send changes, retrieve the server version, preview Floorplans, and test Dynamic Logic without breaking focus.

From metadata to a working request

Explore Generic APIs and Integration Points, follow links between OHI APIs and object definitions, then open the result in API Search or API Client with metadata-aware assistance.

From one request to repeatable automation

Move useful searches and API calls into Runner. Runner tabs keep multiple independent setups open, while a shared execution lock allows only one tab to run at a time. A source is optional: requests without source-field references run once. Add one CSV, JSON, or managed-auth Source Request when row data is needed. Source Request preview retrieves at most 20 rows without pagination, reports the query total, can be stopped, and refreshes every page only for a real run. A 204 No Content preview reports that no lines were found; if a real run refresh finds no lines, Runner stops before sending requests without showing an error. For request sources, Results shows a dedicated Acquiring Lines result with HTTP, timing, paging, row-count, response, and error details; file sources omit it. Each Run opens a timestamped result box and collapses the previous box so recent runs can be reopened and compared. API Search can create a new Runner tab using its JSON search payload as a first-page Source Request. Runner results distinguish request-execution totals from source-line totals: a source line passes only when every request for it passes, and fails when any request fails.

New to Runner? Open the ? control in any major Runner block for the guided quick start, field-by-field examples, result-reading guide, and troubleshooting. The extractor and assertion reference explains every JSON field, Kind, operator, and copyable recipe. Documentation search covers full explanations, option tables, and code examples—not only page titles. The same durable material is available in the Runner reference.

Runner scripts

Each Runner request can have a pre-request hook and a post-response hook. Hooks run for every source row. Pre-request hooks can edit the rendered relative URL, query parameters, non-auth headers, body, content type, and accept header; post-response hooks can inspect the status, redacted response headers, body, and duration. Both use row-scoped runner.variables, which later steps reference with {{vars.name}}, and runner.assert(condition, message) records a visible check. While editing, type runner or runner. (or press Ctrl+Space) to complete the phase-specific Runner API, previewed source fields, extractor values, and variables created by earlier scripts.

// Pre-request: transform the request and share a value with a later step.
runner.request.url = `/api/generic/policies/${runner.row.policyId}`;
runner.variables.set("policyRef", runner.row.policyId);

// Post-response: make a response check.
runner.assert(runner.response.status === 201, "Expected a create response");

To inspect response items, use a post-response hook such as const body = JSON.parse(runner.response.body); for (const item of body.items || []) runner.log(item.code);. Open Runner Output displays these lines in the dedicated OHI Studio - Runner output channel; do not log sensitive values. Runner can also reopen a completed run-*.json result export to restore its request definitions and options.

Hooks run in a QuickJS sandbox with a 250 ms CPU limit and 32 KiB source limit. They have no Node.js, filesystem, network, VS Code, module, or credential access. A script error, timeout, invalid request mutation, or failed assertion fails the step and follows Continue this row after failure. Use the Extractors and Assertions JSON arrays for standard work; reserve scripts for custom transformations and validation. Review imported scripts and never store secrets in scripts or variables. Open Runner Script API from either script editor for every object and function with examples, or see the Runner reference.

From a question to an agent-assisted OHI workflow

Ask an MCP-connected agent to discover the right target and API, inspect metadata or logs, prepare a test, or carry out an explicitly authorized change—without exposing stored credentials or changing the environment selected in VS Code.

Get started

1. Run OHI: Manage Environments and add an OHI environment and application.
2. Select the active environment and application from the VS Code status bar.
3. Start the MCP server for agent workflows, open OHI: PostOHffice, or work from a Dynamic Logic or Floorplan file.

VS Code 1.85 or newer and access to an OHI environment are required.

Credentials stay outside the workspace

Tokens, OAuth passwords, and client secrets use VS Code Secret Storage. Exported environments omit credentials, PostOHffice does not receive stored authentication values, and OHI landing pages use their own browser sign-in.

MCP access is configured per environment, with read-only or read/write scope and additional controls for write-capable tools.

Learn more

Open OHI: Documentation from the Command Palette and start from the task you want to complete. The in-extension manual separates guided tutorials, focused how-to guides, factual reference, concepts, and symptom-led troubleshooting; it shows one page at a time with breadcrumbs, related topics, previous/next navigation, full-content search, and copyable examples. A ? on an OHI Studio screen or complex field opens its exact documented topic directly, falling back to the containing guide only when no focused topic exists.

• Install from the Visual Studio Marketplace
• Browse the detailed reference
• Report an issue or suggest a feature

License

MIT