# One U Education (万友教育) > One U Education (万友教育) is an Australian study-and-migration platform. It > indexes roughly 50,000 CRICOS-registered Australian courses across > hundreds of schools, plus occupations, visa guides, and skill assessment > guides for people studying or migrating to Australia. > > 万友教育是一个澳洲留学与移民信息平台,收录约 5 万门澳洲 CRICOS 注册课程、院校、 > 职业、签证与技能评估指南。 One U Education is a private education agency and information service, not affiliated with the Australian Government and not a Registered Training Organisation (RTO). Every course carries the CRICOS code and RTO details of the institution that actually delivers it. Occupation MCP queries accept the internal `occupation.occupationCode` from search results or an occupation URL/slug. `occupation.anzsco` is a shared classification code; `dataScope` identifies shared statistics/rules. Exact internal matches take precedence. Missing optional identity/name fields are omitted, and names retain their source language/version. EOI tools return English prose; global rankings describe ANZSCO groups rather than unique internal occupations. Visa queries include mandatory assessment exemptions; failed state data sources return `partial`, not empty successful results. ## Site conventions - Production domain: `https://www.oneuedu.com` - Locales: `zh-CN` (default, no path prefix), `en-US`, `zh-TW` — every non-default locale is prefixed, e.g. `https://www.oneuedu.com/en-US/schools`. - Course detail URL pattern: `/courses/-`, e.g. `/courses/master-of-nursing-006254G`. Any `/courses/-` 301-redirects to the canonical slug, so linking with just the CRICOS code suffix (and any slug prefix) still resolves. - Any request with an `Accept: text/markdown` header receives a Markdown rendition of the page instead of HTML — this applies site-wide, including course, school, and search-result pages. ## Course pages - URL pattern: `/courses/-` - Each course page embeds schema.org `Course` JSON-LD in the page ``. - Requesting the page with `Accept: text/markdown` returns a clean Markdown rendition (no navigation chrome), suitable for extraction by agents. - Key fields available on the page and in the JSON-LD: tuition fee (varies by student location — onshore vs offshore), course duration, intake dates, minimum IELTS score, campus/city, pathway occupations (for skilled migration relevance), and the provider's official visa risk/evidence level. ## Course search - Base URL: `https://www.oneuedu.com/schools` - Full parameter reference (types, allowed values, redirect behaviour): `https://www.oneuedu.com/llms.txt` (this file) and `docs/course-search-url-parameters.md` in the site's source repository. - Multi-value parameters are passed as **repeated keys**, not brackets: `state=VIC&de=master&de=bachelor` (the older `state[]=VIC` form still works but 301-redirects to the repeated-key form). - Always pass `studentLocation` explicitly. If omitted, the server falls back to a saved per-visitor preference (or a default), so the same URL can render different fees/results for different visitors — bad for anything you plan to cite or cache. - `mindur` and `isPte` are not reliable: `mindur` currently has a parsing bug that drops the lower bound of the course-length filter, and `isPte` is accepted by the URL builder but never read back by the search page. Avoid depending on either. ### Parameter quick reference | Parameter | Type | Meaning | | --- | --- | --- | | `k` | string | Free-text keyword (course name, Chinese name, or CRICOS code) | | `state` | string (repeatable) | Australian state/territory — only the first value is used | | `city` | string (repeatable) | City — only the first value is used; needs a matching `state` | | `de` | string (repeatable) | Broad degree level | | `del` | string (repeatable) | Exact degree level (overrides `de` when present) | | `sub` | string (repeatable) | Broad subject/field code | | `studentLocation` | string | `onshore` or `offshore` — always pass this explicitly | | `minf` / `maxf` | number | Tuition fee range (AUD, whole numbers) | | `miniel` / `maxiel` | number | IELTS overall band range (0–9, step 0.5) | | `mindur` / `maxdur` | number | Course length range in years (0–7) — see caveat above for `mindur` | | `fu` | string | Institution type — see values below | | `ed` | string | Education type — see values below | | `sc` | string | Restrict to one school's CRICOS code | | `isR` | 0/1 | Regional-area courses only (`1`) or non-regional only (`0`) | | `isC` | 0/1 | `0` = still accepting int'l students (default), `1` = closed to new int'l students | | `pkc` | 0/1 | Packaged courses only (`1`) | | `sf` / `sv` | string / number | Sort field / sort direction — must be passed together | | `p` | number | Page number (9 results per page) | `state` values: `VIC` Victoria, `NSW` New South Wales, `NT` Northern Territory, `ACT` Australian Capital Territory, `QLD` Queensland, `SA` South Australia, `WA` Western Australia, `TAS` Tasmania. `de` values: `master`, `bachelor`, `diploma`, `k12`, `other`. `sub` values (ASCED broad field codes): `01` Natural and Physical Sciences, `02` Information Technology, `03` Engineering and Related Technologies, `04` Architecture and Building, `05` Agriculture/Environmental and Related Studies, `06` Health, `07` Education, `08` Management and Commerce, `09` Society and Culture, `10` Creative Arts, `11` Food/Hospitality and Personal Services, `12` Mixed Field Programmes. `studentLocation` values: `onshore` (student is inside Australia), `offshore` (student is outside Australia) — affects both the result set and the fee figures shown. `fu` values: `Government`, `Private`. `ed` values: `higherEducation`, `vocationalEducation`. `sf` values: `fee`, `totalFee`, `length`. `sv` values: `1` (ascending), `-1` (descending). `isC` values: `0` (still open to international students), `1` (closed to new international students). `p` values: positive integers; results beyond the last page return an empty list rather than an error. ### Example URLs ```text https://www.oneuedu.com/schools?k=nursing&state=VIC&city=Melbourne&de=master&studentLocation=offshore&isC=0&p=1 https://www.oneuedu.com/schools?state=NSW&sub=02&ed=higherEducation&studentLocation=onshore&isC=0&sf=fee&sv=1&p=1 https://www.oneuedu.com/en-US/schools?k=early%20childhood&de=diploma&studentLocation=offshore&isC=0&p=1 ``` ## Occupations & skills assessment Every ANZSCO skilled occupation has a page at `https://www.oneuedu.com/australian-skill-assessment/-` (e.g. `/australian-skill-assessment/nurse-practitioner-254411`; any `/australian-skill-assessment/-` redirects to the canonical slug). Each page carries schema.org `Occupation` JSON-LD and covers: description and tasks, skill level, occupation lists (MLTSSL / STSOL / ROL / CSOL), shortage status by state, assessing authority and the skills-assessment pathways (processing times, requirements), visa options with eligibility, subclass 189 invitation rounds and points, state/territory nomination programs and their requirements, DAMA options, recommended courses and Q&A. Skills-assessment qualification candidates retain the precise evidence limitation. An occupation-level unknown does not mean every listed candidate is unverified, and a listed qualification does not establish an individual's eligibility. - [Occupation search](https://www.oneuedu.com/australian-skill-assessment/all): `?keyword=&type=&page=`; 15 results per page. - [Occupation lists explained](https://www.oneuedu.com/australian-skill-assessment/lists): What MLTSSL, STSOL, ROL and CSOL mean and which visas they unlock. - [State nomination quotas](https://www.oneuedu.com/australian-skill-assessment/state-nomination-quotas): Latest state/territory nomination allocations. - [State nomination guide and self-check](https://www.oneuedu.com/tools/state-nomination): 190, 491 and DAMA status for every state, per-stream requirements, occupation lists, remaining places and an online self-check (`/tools/state-nomination/`). ## MCP server A read-only Model Context Protocol server exposes the same course, school, occupation and skilled-migration data (SkillSelect EOI backlog, 189 invitation history and wait estimates, state nomination, visa charges, processing times, student visa grant rates and skills assessment rules), plus Registered Training Organisation (RTO) registration and regulatory records from training.gov.au, and the partner agent price list (prices need the partner unlock code), programmatically: - Endpoint: `https://mcp.oneuedu.com/mcp` - Transport: Streamable HTTP. Public lookups need no authentication; EOI history requires an OAuth account connection when optional authentication is enabled. - Server card (MCP Server Card extension, SEP-2127): `https://mcp.oneuedu.com/mcp/server-card`, listed in the AI Catalog `https://www.oneuedu.com/.well-known/ai-catalog.json`. The older site descriptor `https://www.oneuedu.com/.well-known/mcp.json` stays available. - Tools: `import_feedback_attachments`, `prepare_feedback_upload`, `submit_feedback`, `list_my_feedback`, `get_my_feedback`, `search_rpl_qualifications`, `get_rpl_qualification`, `search_short_courses`, `get_short_course`, `search_graduate_jobs`, `get_graduate_job`, `search_courses`, `search_course_lists`, `get_course_list`, `get_course`, `get_course_pathways`, `get_course_packages`, `get_similar_courses`, `get_school_campuses`, `get_school`, `search_occupations`, `get_base_occupation`, `get_occupation_visas`, `get_occupation_courses`, `get_occupation_labour_agreements`, `get_occupation`, `list_search_options`, `get_skill_assessment_rules`, `get_eoi`, `get_189_invitations`, `estimate_189_invitation`, `get_state_nomination`, `get_state_nomination_detail`, `get_state_nomination_occupations`, `get_state_dama_agreements`, `query_dama_occupations`, `get_visa_fees`, `get_visa_processing_times`, `get_student_visa_grant_rates`, `search_visa_conditions`, `get_visa_condition`, `list_visa_case_filters`, `search_visa_cases`, `get_visa_case`, `search_postcodes`, `get_postcode`, `search_rtos`, `get_rto`, `get_rto_regulatory_actions`, `get_agent_pricing`, `list_advisors` - Student visa grant rates: `get_student_visa_grant_rates` returns BP0015 decision counts, rates and sample notices for six optional filters (common short names such as VET, ELICOS or offshore work), with an optional breakdown. Rates are aggregate statistics, not an individual applicant’s chance of a grant. Cite the decision period, snapshot and official source. The [student visa grant rates pages](https://www.oneuedu.com/tools/student-visa-grant-rates) show the website defaults. - RPL issuing schools: `get_rpl_qualification` keeps `includeProviderIdentity=false` by default. Set it to true when identifying the school behind an offering or quote; the public endpoint requires account authorization. Ordinary course and price queries remain public. - Advisor directory: `list_advisors` returns official consultation details and contact links as text, without custom cards or avatar URLs. - Resource: `oneuedu://search-options` (the company-service toolkit is served separately) - Prompt: `find_courses_for_student` - Every data query result includes an `attribution` block (source, citation requirement, terms). A per-client daily quota applies; the endpoint is for answering individual user questions, not bulk extraction. ### Website page → MCP tool Data pages end with a short "data interface" note naming the tool and the page's own arguments, for example `get_course(course="006254G")`. The same mapping: | Website page | MCP tool | Argument from the page | |---|---|---| | `/labour-agreements` (filtered by occupation) | `get_occupation_labour_agreements` | `occupation`: the filtered occupation | | `/courses/-` | `get_course` | `course`: CRICOS course code | | `/schools/-` | `get_school` | `school`: CRICOS provider code | | `/visa/dama/` | `query_dama_occupations` | `agreementId`: the region's agreement | | `/tools/eoi-data/...` | `get_eoi` | `occupation`, `visaSubclass` | | `/australian-skill-assessment/` | `get_occupation` | `occupation`, `sections` | | `/rto/` | `get_rto` | `rto`: RTO code | | `/tools/state-nomination/` | `get_state_nomination` | `state`: ACT, NSW, NT, QLD, SA, TAS, VIC or WA | | `/tools/student-visa-grant-rates/...` | `get_student_visa_grant_rates` | `country` or `sector` on filtered pages | | `/rpl/` | `get_rpl_qualification` | `slug`: the page slug | | `/visa/processing-time` | `get_visa_processing_times` | `visaSubclass` | One tool call returns what a page shows across several screens or result pages, such as every employer with a labour agreement for one occupation. ## Other key pages - [School pages](https://www.oneuedu.com/schools/-): School profile, campuses, and course list for one institution. - [Skill assessment guides](https://www.oneuedu.com/australian-skill-assessment/-): Occupation-by-occupation skills assessment requirements for migration. - [Visa guides](https://www.oneuedu.com/visa): Hub page linking to guides for individual Australian visa subclasses. - [Latest visa cases](https://www.oneuedu.com/latest-visa-cases): Real, anonymised visa case outcomes and processing times. - [News](https://www.oneuedu.com/news): Study and migration policy news. - [Sitemap](https://www.oneuedu.com/sitemap.xml): Full URL index of the site, including all locales. ## Usage This content is free to cite, quote, and summarise with attribution and a link back to the source page. Tuition fees and other figures are indicative and change over time — always confirm current fees and intake dates with the listed course provider before advising a student. Please do not scrape faster than 1 request per second; use the MCP server for bounded lookups that answer individual questions. Bulk extraction, redistribution and AI training use are not permitted. Questions or takedown requests: `https://www.oneuedu.com`. Browse published course lists or search by title or course keyword with `search_course_lists`. Pass the public ID to `get_course_list` to read courses in website order, campuses, advisor notes and the website link. Use the CRICOS course code with `get_course` for current fees, English requirements and intakes. Successful visa cases: choose one subclass (500/600/485/820), call `list_visa_case_filters` for valid filter values, then `search_visa_cases` and `get_visa_case`. Public, anonymised backgrounds and website links; no Offer cases, keyword search, similarity matching or approval predictions. Supports zh-CN, zh-TW and en-US. - Occupation MCP routing: use the standalone topic tools for single questions. `get_occupation` requires explicit `sections` (basic, visas, assessment, invitations, eoi, stateNomination, courses, labourAgreements). get_occupation returns English prose for selected topics only, preserves exact identity and source links, and explains missing data and failures. Other standalone topic contracts are unchanged. Occupation applicability is not personal eligibility. State nomination tools: get_state_nomination provides website policy overviews and occupation matching. Program summaries include lastVerifiedAt when the upstream record provides it. get_state_nomination_detail requires state and nonempty sections, returns selected blocks in full, and requires visaSubclass plus an exact directory program for programRules (whole-state upstream fetch, one-project output). get_state_nomination_occupations returns the complete program directory or a single list at 40 entries per page. query_dama_occupations lists regional occupation entries. Use the returned agreementId, itemId and releaseId to retrieve the occupation’s conditions, exceptions and agreement sources in the same release. get_state_dama_agreements requires employerState and returns 20 employer records per page including agreementType; employer location does not define regional DAMA coverage. Use the returned source dates and links when explaining policy; request the allocations section for state quotas. Quota observations retain state/subclass dates separately from cumulative nomination reporting dates; unknown figures are not zero. ### Postcode and regional-area lookup `search_postcodes` searches postcode and English/Chinese locality names (20 results by default, at most 50 per page). `get_postcode` accepts a four-digit string, preserving leading zeroes, and returns all matching states/localities and Category 2/3 classifications. School results match campus address text and are deduplicated by school code; they are not distance-ranked or guaranteed open. Regional status does not establish personal visa eligibility. Visa conditions: `search_visa_conditions` searches condition numbers, titles and explanations with pagination by condition number; `get_visa_condition` returns all official explanations for one four-digit number in full Markdown. Supports en-US, zh-CN and zh-TW, with sourceUrl, retrievedAt and contentVersion. English is the official text; Chinese translations are not official. The snapshot covers the Home Affairs public conditions list, not every possible statutory condition. Multiple source records can share a number and must not be silently merged or selected as personally applicable. This lookup does not establish which conditions apply to an individual visa. EOI disclosure: get_eoi returns English prose. view=summary (default) gives a short overview and optional points comparison; view=history returns at most six monthly totals; view=points or states requires occupation and month and returns at most fifteen categories per page. Every view allows the latest published month and preceding two calendar months anonymously. Earlier months require a connected One U Education account. History uses fromMonth/toMonth, including the suggested toMonth for earlier pages; distributions use page. Page sizes and dataset versions are managed internally. Continue only when needed to answer the user. Counts below twenty are suppressed, missing data is distinct from zero, and state totals overlap. Links point to the corresponding One U Education EOI page. get_occupation provides a three-stream EOI overview and directs detailed queries to get_eoi. School/course MCP details: get_school and get_course return English prose with existing website links; get_school, search_courses and get_course give the provider's official logo as logoUrl (HTTPS, or null when no verified logo is recorded; never a campus photo); locale selects the website-link language only. Basic information and P1/P2/P3 processing priority are public. Provider evidence is read only when includeEvidence=true; the public endpoint then requires website OAuth. search_courses evidenceLevel (1-3) limits results to providers at those evidence levels, also behind website OAuth; each returned provider is checked against its current level record and shown with its capture date. Default detail queries do not read evidence, including for signed-in accounts. Page URLs are source citations; protected provider data queries require an account connection. passportCountry and studentType have no defaults: confirm the passport country and the Document Checklist Tool category before querying documentary outcomes; currently only 01 (None of the above / all other students) is supported. Country levels, provider levels, processing priority and One U reference risk are distinct. Documentary outcomes retain capture dates and observed/derived/unknown status; missing data never means no requirements. Do not infer hidden levels. Public data-query MCP results are English natural-language text. Course detail defaults to core; explicit sections retain complete on-demand topics and version-bound pagination. School evidence requires includeEvidence=true and website OAuth; basic queries do not automatically read protected data. The company-service structured endpoint uses a fixed API Key and does not apply website-user data restrictions. submit_feedback is available with or without an account connection. Signed-in submissions are associated with the verified account. Abuse prevention applies. It saves user-requested feedback for review and returns a receipt; it does not edit official records. prepare_feedback_upload requires an account connection and returns parameters for direct CRM upload plus text instructions: HTTP PUT, X-Upload-Token, application/octet-stream raw bytes, a curl example, and confirmation of HTTP 200 / matching attachment ID / ready before submit_feedback. Limits are 10 files, 20 MiB per file and 50 MiB combined per feedback including later additions. A client without local file or HTTP access must explain that limitation rather than claim an upload or silently omit the attachments. Supported chat-file hosts can use import_feedback_attachments instead. Write feedback content so a reviewer can understand it without the conversation: preserve the known school/course and narrower scope, concrete issue or correction, applicable dates/conditions, impact and shareable evidence. Use existing context first; ask only about consequential missing facts, and accept useful leads with explicit unknowns. Keep user claims distinct from verified facts; do not invent causes, dates or evidence, generalize a course issue to the whole school, or use the submission year as the applicable year by default. ### Account-connected application cards (where enabled) Application access requires an account connection with applications:manage and a host enrolled for hidden credential delivery. open_application_form returns optional saved summary data from an authorized application read; summaryStatus distinguishes unavailable summaries. File saving counts do not imply material completeness or review. Forms credentials remain hidden in the card. Opening a form does not submit to an application specialist or a school. Company resource tools use a separate signed host endpoint, not the public MCP tool package. Authorized hosts can submit, read and withdraw resource review requests through resource_review_submit, resource_review_get and resource_review_withdraw. REVIEW_SUBMITTED means awaiting review rather than publicly published. Approval and rejection occur only on the internal website. Feedback discussions: list_my_feedback and get_my_feedback read only the current account’s signed-in reports and public messages. submit_feedback with feedback_id appends to that same ticket; anonymous reports remain one-way. Use one request_id per write and preserve it on uncertain retries. Notifications for new signed-in reports default to the account email; contact_email overrides only when explicitly supplied. Anonymous email is optional. notifications_enabled=false opts out. Staff replies and resolution generate template notices; anonymous reports receive resolution notices only. notification_locale supports en, zh-CN and zh-TW (default en). Course major directions: get_course with sections=["majors"] returns published directions within the selected CRICOS course, source facts and selectionRequired (true/false/null). Having options does not imply compulsory selection or availability at every campus/intake. Use majors.page/pageSize and preserve contentVersion for continuation; no published data is not proof that there are no majors. Major queries bypass the MCP edge cache so subsequent reads can observe committed maintenance changes. ### Account-connected migration planning (where enabled) With planning:read, get_planning_input_options reads Forms input choices and coarse categories; search_planning_occupations discovers the complete Forms catalogue using optional q/category/cursor/limit, without a dossier or writes. Preserve internal identities, classification-level present/absent/unavailable source coverage and actual consumed snapshots; coverage does not establish predictability or personal eligibility. Old cursors fail on source/query changes. Existing search_occupations remains general public catalogue search. The existing MCP may expose list_planning_dossiers, create_planning_dossier, get_planning_dossier, update_planning_profile, update_planning_settings, create_planning_calculation, list_planning_calculations, get_planning_calculation, get_planning_schema_index and get_planning_schema_definition when its Forms planning service is configured. Requires explicit planning:read, planning:write or planning:calculate consent in addition to mcp:read. Client models interpret user text/files into structured confirmed facts; clarify ambiguity before saving. Use get_planning_schema_index without index_ref for the root, use root topic index_ref links directly, then follow references for nested records, dictionary values and variants. Navigation references belong to one schema name/version. Use get_planning_schema_definition with a schema_refs list of 1–20 published references (one also uses a list) to read definitions with shared closed components; invalid or oversized batches fail explicitly; roots for every schema name publish a definition reference. Neither tool accepts who/group/view. Only the current version is served. Update discovery uses a strict envelope and lightweight changes; complete generated and backend validation still run. Discover unfamiliar field/record shapes and reuse included definitions/components at the same schema name/version, without mixing profile output with profile_patch input. Schema reads are not required before every write; save supported facts without completing unrelated topics. PATCH uses expected_revision; arrays replace complete arrays. Creation uses request_id for idempotency. Calculations return persistent IDs asynchronously. Fixed calculation_scope=existing_189|future is independent from calculation mode. MCP accepts only ordinary submissions (the default); heuristic and new comparison calculations requiring heuristic are unavailable through MCP. Saved heuristic results remain readable. Website/API calculation modes are unchanged. Full planning starts existing_189, shows available routes, and uses its ready route_summary over all 189 ordinary/forecast routes to decide whether a usable conditional PR date exists. With a date, stop automatically; without one, continue future within authorization, unless explicitly 189-only. Errors, failed forecast and unavailable/unread decisions are not normal no-date results; disclose partial scope. Before linked future work, verify the current dossier revision equals first input_revision, then use continuation_of, the same revision/mode and a distinct stable request_id. Do not bypass revision/version/continuation errors by dropping the link or changing keys; query saved progress/results, or recover by request_id after timeout. get_planning_calculation section=list returns a lightweight overview; section=reading returns selected explanation details; section=result returns raw pages. Start with list to locate candidates and scope. A directions-only question may end with that overview; for planning, next steps or a substantive result explanation, read the necessary conditions, preparations, dependencies, sources, time and costs of the most relevant 1–2 candidates before replying, without asking permission to inspect them. Selection for explanation does not save goals or authorize recalculation. Explain common preparation once, distinguish occupations from visa endpoints, and preserve separate ordinary/forecast references, conditions and coverage without adding counts. When the user names a route or condition, directly use reading_ref.section and reading_ref.path as section and result_path, then the relevant reading_sections, without asking permission again. Before detailed claims, finish the supporting source, condition and dependency pages/references and retain source limitations. Read preparations, time and cost when needed for the requested answer, without opening every section or route; When returned condition_sources/sources references explicitly identify the selected route’s conditions, read those details first to obtain condition text and sources together, then supplement reading only for missing fields. Reuse already-read corresponding condition fields across reading/result; matching paths alone do not establish equivalence. Keep the default page size for condition/source objects; small limits suit rich route collections. Follow unread references/pages needed for the claim. Once source text, URL and applicability limits suffice, stop without chasing optional display/summary copies; read display when it holds necessary evidence or limits absent from the source text. A scalar result_value_too_large (413) cannot be fixed by lowering limit; if necessary evidence remains unread, state the limitation and answer only from what was read. detail_path belongs to raw result, not reading. Keep the same dossier_id/calculation_id and clear cursor whenever section or path changes. Start at result_path="", use root topic/entry paths, and retain next_cursor. Read all pages/references of each relevant selected section before claiming complete coverage; explanations need not rebuild every raw result graph. Reading routes carry their conditions and statistics. The result_list Schema describes the list projection; result_reading describes expanded detail content and field meanings, separate from full-reading record tables; inline subtrees are complete within their projection; a complete list does not contain the full explanation. Default page limit is 100; narrow rich route collections explicitly. Questions include current revision, profile schema version, authoritative blocking_gaps and local scope_gaps with their own preferences/profile references. Discover settings fields with schema name=preferences; applicant and non-recursive spouse preferences are independent. A skip preserves the saved answer; explicit undecided clears it with null. Forms restricts dependent actions for unresolved material preferences while other directions can continue; scope_gaps are separate from global blockers, ineligibility and resource-limited partial search. Preserve these explanations in saved results; fetch before a requested calculation or when the next reply needs a fresh backend question, not after every save. Empty blockers permit other optional facts to remain unknown. Status performs up to ten seconds of waiting time on the same pending calculation and returns its actual state with model-visible retry_after_seconds; continue that ID without resubmitting. Inline values, including null, remain complete; no routes or conditions are truncated. Preserve partial completeness, conditions and sources. Website anonymous dossiers and MCP account dossiers/quotas are independent. VIP is unavailable through MCP. Fixed company-service keys do not grant private dossier access. Planning results use the validated Forms structured envelope with explanatory JSON text. A status-only question still requires current status; unread results do not mean an unfinished calculation. Calculation success is not delivery of a requested plan: keep polling the same task using retry_after_seconds, and after success read and explain available routes even without PR dates or with partial search. A progress follow-up continues an undelivered plan unless explicitly status-only, paused or cancelled. If waiting must end, verify a supported authorized follow-up or state that none was scheduled. Preserve successful ordinary results when forecast fails. The full cross-tool workflow is maintained in the OneU planning skill. Planning follows create → update structured facts → submit (needs_information or queued) → read fixed results. Study, work and confirmed occupational goals remain independent. Nullable empty strings/arrays become null; preserve valid zero and false. Current and expected passport countries are separate. Read the authoritative versioned Schema and retain information gaps, result reading views, dependencies and partial completeness. No backend LLM/OCR, original-file storage or saved model explanations; saved business records are not cleared for inactivity and remain subject to active deletion. Course search empty results: no_data means no candidate providers were returned for evidence-level verification and coverage is unknown; no_match means no records matched the evaluated query. Neither proves that no such schools exist. System and authorisation failures remain errors. News feedback: target_type=news requires an account and news_category=school|visa. Optional proposed_school_code/name are suggestions; internal staff confirm one main school per news item. Existing ordinary anonymous feedback stays available. Original files: import_feedback_attachments uses supported host file APIs; prepare_feedback_upload provides a fallback raw-byte PUT capability for local clients. Both require login. Pass ready attachment_ids to submit_feedback. Maximum 10 files per feedback, 20 MiB each and 50 MiB combined including later additions; keep original PDF/Office documents instead of parsing brochures. Reuse request_id on uncertain retries; file import is not publication. Staff rewrite the feedback body for public readers, remove body signatures and contact details, and publish through the internal OneULink CMS workflow. Course payment schedules: explicitly request get_course sections=["fees"] with the required audience and years. Each audience/year may have one optional schedule; payment amounts are AUD cents internally and render in AUD. Preserve order, deposit/balance wording, billing units, free items and unknown amounts. Schedules do not depend on campuses. sections=["campuses"] and course-scoped get_school_campuses also include general and campus teaching arrangements by year; fields inherit separately, and an explicit empty location list stops inheritance. These managed sections bypass the MCP edge cache. Packaged-course feedback: submit_feedback(target_type=course_package) accepts text and optional evidence links, with no attachments or file requirement. Use the school code/name for the target and describe component courses, sequence, onshore/offshore package totals and conditions in content. Package query totalTuitionAud is offshore; onshoreTotalTuitionAud is separate and never falls back to offshore. Planning single-goal detail: get_planning_calculation section=goal requires goal_path copied from the raw goal catalogue; result_path is relative to that target and cursor must reset on target change. Discover fields with result_goal. Read canonical /records/pr_forecast/{index} and /records/processing_statistic/{index} references with section=reading in the same calculation and reuse their complete bodies across routes. Preserve distinct targets, unknown conclusions and unread pages. Goal-page transport: kind=shared entries use record_id to reference complete values in the records map; retain each distinct goal_path and all unknown/failed targets. Read selected goal dependencies first, reuse included prediction/statistic evidence, and group independent known reads in one tool round. Missing records remain unread. Course class offerings (ONEU-66): get_course with sections=["offerings"] reads morning/evening classes by exact year. Use offerings.year and offerings.offeringId (20-character string Unique ID), then offerings.quote={audience,weeks,startDate,bookingType} for a fresh server-calculated quote. Amounts are AUD integer cents. The base price and conditional promotions remain distinct, and school confirmation is required for promotion eligibility. Use page/pageSize/contentVersion for continuation. The offerings topic bypasses edge caching. A year using offeringMode=offerings requires class-specific fees and dates. State government school programs: courses run by the eight state education departments (provider codes ACT 00643J, NSW 00588M, NT 00780M, QLD 00608A, SA 00018A, TAS 03352G, VIC 00861K, WA 01723A; for example VIC 019048F) return state-wide data in get_course: core summarises tuition by year level, the earliest term still open with its deadline, the first invoice and entry rules; fees lists every published charge by payee, homestay, the first-invoice and first-year estimates with their basis, and labels the CRICOS register figure as a whole-course estimate rather than annual tuition; requirements gives visa age limits, English by year band, academic results and guardianship; intakes lists state terms with closed/earliest/open status. Optional fees.stateSchool={yearLevel,startTerm,startYear,accommodation} narrows the estimate (startTerm and startYear together). get_school for those provider codes adds the program summary; get_visa_fees for subclass 500 adds school-student age limits and financial capacity amounts. Unpublished items stay "not published" or "not confirmed"; no deadline is invented for states that do not publish one.