← Back to selected work

02 · Workiva MCP & API layer

Workiva MCP Server

Authenticated API tools so AI workflows read and update Workiva spreadsheets and documents—without manual browser work.

Evidence map

29 automated tests across nine acceptance areas

The integrated lab exercises deployment, observation, and recovery behavior.

  1. Containerized service
  2. Authenticated API
  3. Tool integration
  4. Failure taxonomy
  5. Acceptance harness
Boundary

Reproducible lab evidence—not a live client production deployment.

Constraint
Browser-only automation breaks on canvas grids, iframe boundaries, and session expiry; regulated reporting needs structured reads, writes, and readback.
Built
MCP server exposing 20+ Workiva REST tools—values, formulas, formats, destination links, Wdata queries, file exports, and async job polling—with engagement-scoped credentials and production error handling. Paired with a containerized deployment lab (29 automated tests, nine acceptance areas) that proves clean-clone deploy, auth, and failure recovery.
Role
Built and maintain the integration; designed the deployment shape, service harness, and operating runbooks.
Human boundary
API-first for structured data; browser capture only for UI-only proof. The lab proves reproducibility—it is not represented as a live client production tenant.
Evidence
29 automated tests across nine acceptance areas; live readback after scoped writes; binary-safe export verification.
Scope
Architecture, tests, and operating discipline are public; client data and credentials are not.
Get in touch