# Service Portal

> Build and operate guided service workflows from one platform

Published workflow count: 0

## Site identity
- Brand: Service Portal
- Canonical site origin: https://www.feinprov.com

## Core capabilities
- No service capability classification is currently provided by the published profiles.

## Workflow execution
- Prefer the structured public API sequence: read workflow metadata, start a session (the server advances start and automated nodes), submit one confirmed answer for each returned question, then repeat until a resolved terminal result.
- Browser fallback: open the canonical workflow URL in an interactive browser or browser-capable agent and follow the visible prompts in order.
- Read the workflow title, description, profile, limitations, and sources before acting on its result.
- Provide only information supplied or confirmed by the user. Ask for missing information instead of inventing values or assuming eligibility, pricing, availability, or location coverage.
- Ask for user confirmation before submitting personal, contact, payment, booking, or lead information.
- Treat a workflow as complete only when it displays a final result, request status, or confirmation.

## Decision-tree operating protocol
- A workflow is a server-controlled state machine. The returned current node is authoritative; automated nodes are advanced by the server, so never submit answers for them or guess a branch.
- Start: POST `/api/public/session/start` with `{"tree_id": <published tree id>}` and optionally `initial_context` supplied by the user.
- Read: GET `/api/public/session/{session_id}/status` for the current node, or GET `/api/public/session/{session_id}/projection` for an auditable timeline and edge decisions.
- Interpret: inspect the node `type`, `content`, `metadata`, `answer_type`, validation rules, and widget information before asking for an answer.
- Answer: POST `/api/public/session/{session_id}/answer` with `{"answer": <value>}` only after the user confirms the value. Reuse one `Idempotency-Key` when safely retrying the same answer.
- Continue: use the returned `node`; if the response is asynchronous, poll its returned async-status path before submitting another answer.
- Stop: stop on `completed`, a terminal result/error, an approval wait, a trial limit, or a request for information the user has not supplied.
- Widget meaning is metadata-driven. Supported public workflows may use text, textarea, number, date, boolean, enum, multi-select, object, array, JSON, table, upload, catalogue, commerce, or other approved widgets.
- Compute and Goja-backed nodes run on the server. Never expose or execute their source; use only returned output and published explanation.

## Public workflow API
- `GET /api/public/trees` — list published public workflow summaries.
- `GET /api/public/trees/{tree_id}` — read workflow metadata, SEO profile, schema data, and start preview when published.
- `GET /api/public/trees/{tree_id}/semantic` — read the sanitized, author-declared public steps, decision factors, documented outcomes, and published workflow media without runtime node metadata.
- `POST /api/public/session/start` — create a public session; body requires `tree_id` and accepts optional `initial_context`.
- `GET /api/public/session/{session_id}/status` — read current node and completion state.
- `GET /api/public/session/{session_id}/projection` — read session timeline; supports conditional refresh with `If-None-Match` and may return `304`.
- `POST /api/public/session/{session_id}/answer` — submit one answer; body is `{"answer": <JSON value>}` and may return a synchronous next node or `202` with an async job.
- `GET /api/public/session/{session_id}/async-status?job_id={job_id}` — poll an accepted asynchronous answer.
- `POST /api/public/uploads/presign` — create a short-lived signed upload URL for a file required by the current public workflow.
- `GET /api/public/articles/{article_slug}` — read one published article as informational context.

## Full machine-readable context
- [Full public context](https://www.feinprov.com/llms-full.txt): dynamically generated full public pages, articles, workflows, semantic explanations, formulas, FAQs, and documented outcomes.

## Agent interpretation
- Prefer the canonical workflow URL listed for the matching user intent.
- Use the public source and last-verified fields when deciding how current a fact is.
- Do not treat SEO metadata, schema markup, an article, or an agent manifest as proof that a service is available or that a quote, booking, or lead has been accepted.

## Public decision tools
- Direct tool execution endpoint: https://www.feinprov.com/api/mcp (MCP Streamable HTTP). Use find_workflows to search by user intent, then use get_workflow_brief for a compact published explanation with citation URLs and missing evidence. Start, answer, poll, upload, and complete the selected workflow only when the user wants a personalized result.
- No public decision tools are currently available.

## Public site pages
- [AI Agents](https://www.feinprov.com/ai-agents): Public Service Portal information page.
- [Solutions](https://www.feinprov.com/solutions): Public Service Portal information page.
- [Industries](https://www.feinprov.com/industries): Public Service Portal information page.

## Interpretation and trust
- Only public, published, indexable content is listed here.
- A workflow result may be preliminary and should be confirmed when the workflow states that confirmation is required.
- Business identity, service area, pricing, availability, and contact information should be treated as current only when the public page provides a verification date or official source.
- Private sessions, customer submissions, admin pages, unpublished trees, and noindex pages are not included.

## Official links
- [Service Portal](https://www.feinprov.com/)
- [Privacy policy](https://www.feinprov.com/privacy)