FrameOn — Agentic Project Management

Reads one document or Wiki page in full, by id. Content past 25,000 characters is truncated and the response says so explicitly — when you see that flag, tell the person, instead of reasoning as though you had the whole text. Use it after search_documents or list_documents_brief pointed at the right page. It is not a way to browse: reading the Wiki page by page burns the context you need for the actual work.

Lists every task YOU own or created that is still an incomplete record, across all projects of this workspace: missing assignee, start or due date, priority, story points, estimated hours, a parent (for non-epics), or a done task with no hours logged. Each item names its project, its LAB key and exactly what it lacks, plus how to fix each field. These are the same holes the alert engine reports to the team. Work through them with update_task (ids from get_project_team / get_project_structure; ask the person for dates and estimates — never invent them) and log_time. For hours on a done task: reopen to in_progress, log, close again. Read-only and paginated: limit (default 20, max 50) and offset; follow next_offset until it is null. Write responses already show the top five as pending_on_you.

Lists the risk alerts this workspace raised for the project on its own: overdue work, deadlines about to slip, tasks with no movement, no owner or no due date, people overloaded. Read-only, and the fastest way to answer "how is this project doing" with evidence instead of impression — the alerts were computed from the data, not from a summary you wrote. Defaults to active alerts only. Each one carries type, severity, a human message, and the task it points at when it points at one. Two things worth knowing. The list is filtered by what the token's role is allowed to see, so an empty result means "nothing you can see", not always "nothing exists". And you cannot resolve or dismiss an alert through this connector — that stays with a person in the UI, on purpose. Report what is open and let them decide.

Returns everything this project knows about itself as one markdown briefing: decisions taken and why, conventions agreed, what is written in the Wiki, the risks currently open and the work in flight. READ THIS FIRST, before you touch the project's code or answer a question about it. It is the shared memory that replaces the notes file living on one developer's machine — what the last person learned is in here, and it is the difference between continuing the work and rediscovering it. The text is assembled from the workspace's own records and handed to you raw, not summarised by us. Read it yourself and keep what matters. Memories marked confidential are not included, deliberately. The Wiki appears as titles and ids — call get_document_full_content on the ones that matter instead of asking for everything at once. When you finish and have learned something the next person would want, write it back with save_classified_memory.

Returns the project hierarchy: every epic with its children, counts and progress. Pass include_tasks=true when you need the level below stories. This is the ONLY tool that reports per-epic numbers. Call it before create_task and parent the new work under the epic it belongs to — a new top-level task is a decision, not a default. What you see here is how this team actually organises work, which may not follow the epic > story > task > subtask convention: hierarchy in FrameOn is free, any task type may parent any other, and an unusual shape is a choice the team made, not an error for you to correct.

Lists the people on this project with their ids: explicit members when the project restricts visibility, otherwise the active users of the workspace (clients excluded) plus anyone already assigned there. Use the returned id as assignee_id in create_task or update_task — it is required there, so call this first. Never guess an id, and never assign work to someone this tool did not return; when it is unclear who should own a task, ask the person. Read-only, one project per call.

Returns the full script of one workflow: the tools to call, the order to call them in, what to look at in each result, and what to hand back at the end. Read it and then run it yourself with your own tools — these are instructions for you, not a job queued on our side. That is the whole design: the workflow costs the workspace nothing to hand you, and the thinking stays with you. The scripts encode things that are easy to get wrong here and expensive to get wrong twice: that a null figure in a report is missing and not zero, that hours are always logged for the token holder and always land unapproved, that the task hierarchy is a recommendation and never a rule to enforce, and that write steps are proposed to the person first and carried out after they agree. Call list_skills first if you do not know the names.

Returns one task in full: description, dates, assignee, labels, status and parent. Accepts either the human id (LAB-42) or the UUID. Use it when a row from list_tasks needs the detail behind it. Do not loop it over a whole backlog — filter with list_tasks instead, which is one call rather than fifty.

Aggregate counts for one project: totals by status, by type and by priority, plus what is overdue. This answers "how is this project doing" in a single call, instead of paging through list_tasks and counting by hand. It reports the project as a whole — for numbers broken down per epic, get_project_structure is the only tool that has them.

Time report for the project over the last periodDays (default 30, max 365): hours planned versus hours actually logged, the billable split, and the tasks where the two diverge most. Planned counts leaf tasks whose plan overlaps the window; worked counts time entries logged inside it. A null KPI means the other side is missing — it is NOT zero, and reporting it as zero turns "nobody planned this" into "the team overshot by 100%". Use it when someone asks how a period went, before writing any status update. Pair it with get_project_alerts for the risk side and get_task_stats for the delivery side.

Lists this project's documents and Wiki pages with title and id only, newest first. Cheap to call, and the right way to see what exists before reading anything. It is paginated: the response carries total and has_more, and you reach older entries by calling again with page: 2, page: 3, and so on — never treat page 1 as the whole Wiki when has_more is true. To find a specific subject, search_documents is faster than paging through everything. Feed the id you picked into get_document_full_content.

Lists the ready-made workflows this workspace knows how to run: diagnosing a project, reporting on a period, breaking scope into tasks, logging your hours, and inheriting or handing over the project's shared memory. Cheap to call, and worth calling once at the start of a session so you know what exists before improvising a sequence of calls. Each entry carries a name, one line of what it is for, and the tools it uses — so you can tell up front whether this token reaches them. A skill is a script for YOU: it says which tools to call, in what order, and what to deliver. Nothing executes on the FrameOn side, and asking for one costs you nothing but the text. Pick a name and call get_skill.

Lists tasks in one project with optional filters by status, type, priority and free text. Paginated: limit caps at 100 and the response carries has_more and next_page — a backlog you read only to page 1 is a backlog you read wrong. Call this before create_task to find the parent an item belongs under, and before update_task when you do not have the id. Omit project_id once and the error hands you every project this token can reach.

Semantic search over this project Wiki pages and uploaded documents. It matches meaning, not keywords, so give it a question rather than a term. Returns between 1 and 10 excerpts (default 5), never whole pages — when an excerpt looks right, follow up with get_document_full_content. Always search before create_wiki_page: the point is to extend the team memory, not to duplicate it.