Start here: setup, browser sign-in steps, usage examples, and scope. No network.

List normal Chrome/Brave/Edge/Chromium profile names and IDs; no cookie values.

Check live sign-in without opening a browser. Outages are not logged-out states.

Persist non-secret browser selection. Use after the user chooses a profile/source. Clears only this process's cached session, never logs the browser out. Isolated needs no normal-browser extraction; managed uses OpenClaw port 18800.

Open/reuse the selected browser for IE/Microsoft SSO only when the user wants to sign in. Returns immediately. Password/MFA input is human-only in that browser. Does not accept credentials or arbitrary URLs.

Check the active sign-in job; does not open a browser. After expiry use ie_auth_status to detect a sign-in completed later.

Personalized IE Connects dashboard or discovery recommendations.

Search/list events, including your registered upcoming or past events. Logical pagination handles sparse upstream buckets. Use next_offset; scan_exhausted means the bounded search was incomplete, not that no more events exist.

Event details or visible attendee directory. No registration, ticket, or QR credential export.

Find groups/clubs or your memberships; local pagination over at most 1,000 matches.

Read a group's about page or one structured collection; no joins, survey submissions or other writes.

Search the visible IE directory. Respect privacy; an empty list is not proof a person is absent.

Read a profile by returned ID/UID or exact full name; optionally visible connections. Never changes connections.

Student schedule. Date filtering is applied to each upstream page; continue next_offset even if a filtered page is empty. ISO dates; maximum 31-day filter window.

List Connects courses. For LMS readings, assessments and grades use Blackboard tools.

Read a Connects course section using a course ID from connects_courses.

Read feed posts/comments/photos or visible viewer records without posting or changing read state.

Room availability, filters, details, or existing reservations. Cannot book or cancel rooms.

Read jobs, mentorships, marketplace, stores or a fair. No purchases, applications or payments.

Read existing chat data; cannot send messages or change read state.

Read personal activity surfaces. Some return page metadata rather than complete structured records; do not infer completion from them.

Blackboard identity, courses, to-do, activity and message summary (bounded output).

List/search enrolled courses or organizations. Returned IDs are preferred for all other Blackboard tools.

Course details, role, roster/instructors, timetable, tools or progress. Restricted sections remain unavailable.

Browse ROOT/folder children, inspect content, search or list links/attachments. Traverse children for large courses. Does not mark content complete or launch LTI/tests.

Visible grades and EXISTING attempts/feedback. Never starts a test, creates an attempt or submits answers.

Discover assessments or read accessible assessment metadata. Permission-denied authoring fields are not bypassed.

Read course or system announcements. Provide course and announcement ID for detail.

Deadlines, events, calendar definitions, or activity. Events need ISO since/until; range <=28 days. No date arguments uses Blackboard's default to-do window.

Read existing messages/conversations. Never sends messages or marks them read.

Read discussion forums, threads and replies without posting or mutating read state.

Read course groups, group sets and visible membership, with no enrollment changes.

Read recorded attendance or class meetings. Does not check in, fabricate proof or change attendance.

Download an existing attachment (<=100 MiB) to a generated private local file on the MCP machine. No arbitrary URL/path, no overwrite or execution. Returns path, bytes and SHA-256.

Career portal announcements, your calendar, recommended and recent opportunities. Larger nested data may be truncated: use dedicated paginated tools.

Discover live filters, or accepted values for field. Start here before career searches. Autocomplete city/employer needs query (2+ characters); exact labels or IDs avoid ambiguity. No hardcoded career-option lists.

Search opportunities with live typed server-side filters. Use careers_filters first. Values within one field OR together; fields AND together. Ranges accept from/to, exclusions use exclude=true. Defaults to Approved/Application Open; use all_statuses for application history. Follow next_offset with the same query. Dates use Europe/Madrid unless explicitly offset. No applications or favorites are changed.

Job description, requirements, deadline, application methods and source link. Details are projected to career-relevant fields. For long descriptions use section=description and follow next_offset as text_offset (4,000-character chunks). Never starts applications or follows external sites.

Read the career portal's recommendations; these are not filtered search results.

Research employers using their live filter catalog and optional name search.

Employer details, open opportunities, or contacts permitted by IE. Never sends messages or changes target-employer lists.

Read your own career profile, education or recorded interview history. Interviews is not a shared question bank. No profile edits or new interview reports.

Read existing application documents or résumé books. Detail views require an item_id returned by their list. Downloads use careers_download; no uploads or book registrations.

Browse/search career resources or traverse a bounded folder tree. External links are never fetched. Detail needs item_id. Download files with careers_download. Tree is capped at 60 folders; larger results must be narrowed.

Your career calendar or advising availability (up to 31 days), types/offices, and role-dependent events. Empty data can be legitimate; access_denied is not no events. Adviser/type refinements apply locally over the complete date-window response. No bookings or registrations.

Privately download a permitted existing file to a generated filename on the MCP host. Cookie-free allowlisted file transport; 100 MiB/40-second budget. No arbitrary paths, uploads or external URL fetches.

Read local career search presets (shared with iecareers), run one, or inspect portal saved searches. No baseline changes. Show/run require name. Follow next_offset for complete lists.

Validate then save a credential-free LOCAL search preset, only when requested. Shared with the CLI. Does not create portal alerts or schedule/send anything. replace=true overwrites only the named local definition.

One-shot complete search diff: new, changed and no-longer-matching jobs. Default is preview (no write). Only advance_baseline=true commits private local state after a complete scan; reset_baseline handles an explicitly changed query/account. Never schedules or sends notifications. First scan marks all matches added. Persist returned changes before committing if you need reliable downstream delivery; truncated output is not the full change list.