Core Features

Every write or destructive operation is gated behind your explicit approval — the AI cannot approve its own actions:

• Interactive Approvals: A native dialog in Claude Desktop shows the exact endpoint and payload before anything is sent to EPM. Clients without dialog support fall back to a confirmation-token preview flow.
• Read-Only Mode: Block all write operations server-wide with one setting — ideal for production connections.
• Opt-in Security Tools: Role and group assignment tools stay disabled unless explicitly enabled.
• Audit Log: Every tool call is logged locally with status, duration, and inputs (tokens redacted).

Set SERVICE_TYPE=fcc (or both) to enable close-cycle tools alongside the shared EPM toolset:

• Consolidations: Run consolidation, translation, or force-consolidation for any entity/period and monitor the job to completion.
• Journals: List journals across all workflow states, inspect line items, and submit, approve, post, or reject — posting requires approval.
• Ownership & Intercompany: Read ownership percentages and intercompany balances by partner through smart data slices.

• Variance Drill-Down: explain_variance deterministically drills a hierarchy to find what drove a change between two periods, scenarios, or years — all math is computed by the server, never estimated by the model.
• Environment Diff: Sync snapshots from two environments into labeled workspaces and get a drift report of every artifact that differs — the TEST vs PROD check before every deployment.
• Persistent Member Index: Member names, aliases, and hierarchies are indexed locally, so resolution is instant even on large applications.

CaptainMCP allows the AI to securely trigger and monitor Oracle EPM Jobs. This includes:

• Business Rules: Execute calculations, consolidations, and custom Groovy scripts with parameters.
• Data Integrations: Run Data Management load rules to pull data from ERPs.
• Clear/Copy Data: Safely execute data management tasks across slices.

Give your AI the ability to answer complex structural questions about your EPM cubes:

• Hierarchy Traversal: Query parent/child relationships, aliases, and UDAs.
• Substitution Variables: Read and update global or application-level variables (e.g., advancing the `CurMo` variable).
• Smart Lists: Retrieve Smart List definitions and mappings.

AI agents can extract targeted data slices to perform real-time variance analysis or validations:

• Export Data: Run MDX queries or data export jobs to retrieve specific intersections.
• Import Data: Push validated, AI-generated data sets back into the application.