How Plansight connects to the rest of the stack
Audit pages for the systems Plansight integrates with — what they do, how they're wired, what's confirmed, what needs a verify, and what's still pending.
Audit pages for the systems Plansight integrates with — what they do, how they're wired, what's confirmed, what needs a verify, and what's still pending.
Plansight in the middle, the integrations around it. Arrows show data direction. Bright Benefits sits downstream of the Open API as a partner consumer. Click any node to open its page.
Each card opens the full audit. Status reflects how deep the audit currently goes.
Broker management system integration. Bi-directional sync with Vertafore's broker platform. Includes a filterable sync map covering every plan type, product type, and benefit attribute.
AI-powered benefit document parsing (PDF → structured data). Two services: Planfacts API + OpenPlan. 1 confirmed bug, 4 likely-bug items, ~30 small items across 10 benefit types.
Benefits enrollment platform — bi-directional. Push elected plans + employer profile to EN. Pull census directly into community-rated RFPs. OAuth 2.0 with operation-pattern dispatcher. Includes a filterable sync map covering 83 fields across the census + plan-push flows.
Inbound — plan + rate data engine for ACA / community-rated quoting. ~65 plan attributes parsed; rates by rating area; benefit summary PDFs virus-scanned to S3. Working with 6 items — regex fragility, silent PDF fallback worth attention.
Outbound — partner systems call into Plansight. Read-only today (employer list, plan types, plans by type). Bearer token auth, brokerage-isolated, Elasticsearch-backed search, manual OpenAPI spec. Strategic items: deprecation policy, write support, webhooks.
Bright Benefits pulls sold plan data (medical, dental, vision) directly from the Plansight Open API. Activation flow: user creates credentials and a benefit in Plansight, then enables Bright Benefits to consume that data.
Auth, email, AI, AWS, observability, document processing, and more — the third-party services that run Plansight under the hood. Plus deprecated services worth cleaning up.
Each integration deserves a clear, current document of what it does and where the rough edges are. This site is that document.
Each integration page documents: what it does, where it lives in code, the data shape that flows in or out, what's mapped, what's not mapped (intentional vs. missed), what's potentially mapped incorrectly, and a list of small concrete recommendations. Cursor-ready prompts inline so the dev can paste them straight in to investigate.
Plan Facts is the worked example — the deepest audit on the site (10 benefit type sub-pages, API ground-truth confirmations from OpenPlan + Planfacts API, 4 confirmed bugs with Cursor prompts). Other integrations follow the same pattern at appropriate depth.
Each Bitbucket pull triggers a regen. Every page footer stamps the source commit so the version is always clear.
Run python3 scripts/pull-repos.py from the project root — pulls all 4 Bitbucket repos via the configured app password.
Say "regenerate the audit." Claude reads BUILD.md for the procedure and source/analysis-notes.md for previous commit hashes.
Diff against the previous hashes. Re-run the per-integration agents. Update only the pages that changed. Append a row to the version table.
Vercel auto-deploys to integrations.plansight.com. New version is live in under a minute.
Three readers, three different ways into the content.
Per-integration pages have copy-pasteable Cursor prompts inline for any flagged item. Open the page, find the issue, paste the prompt into Cursor.
The dashboard shows status at a glance — green / amber / red badges, item counts, summary text. Skim the cards, click into anything that needs attention.
Source of truth for "what does the integration actually do today, and where are the rough edges?" — for product decisions, customer conversations, roadmap planning.