onplana-project-management

name

onplana-project-management

description

Plan, track, and report on projects, sprints, and portfolios in Onplana (an AI-native project & portfolio management platform, and the Microsoft Project Online successor) over its MCP server. Use when the user asks to create or break down a project, plan sprints/epics/milestones, check status or Earned Value (EVM), run risk analysis, log or review timesheets, or move a proposal through governance. Realistic triggers: 'set up a project in Onplana', 'break this spec into tasks and a sprint', 'what's at risk on the Atlas project?', 'show EVM / status for my portfolio', 'log 3 hours to the migration task'.

Onplana project & portfolio management

Onplana is an AI-native project portfolio management (PPM) platform positioned as the successor to Microsoft Project Online (which retires Sept 30, 2026). Codex reaches it through the remote Onplana MCP server, which exposes 247 tools spanning projects, tasks, dependencies, sprints, epics, milestones, risks, issues, governance/proposals, timesheets & compliance, resource capacity, portfolios, EVM/finance, reports, and workspace documents.

You connect as a real org member (an "agent persona"): you can be assigned tasks, you act under a plan tier and a role, and your writes are attributed to you. Treat Onplana like a teammate's project tool, not a scratchpad.

When to use this skill

Reach for Onplana when the user wants to do project/portfolio work that should persist in their PPM system, for example:

• Stand up a new project and break a brief/spec into tasks, subtasks, sprints, epics, and milestones.
• Check or report status: progress, overdue work, Earned Value (EVM/SPI/CPI), a status digest, or a portfolio rollup.
• Run risk or issue analysis on a project and record findings.
• Log or review timesheets and check compliance.
• Move a proposal through the governance pipeline (Enterprise).
• Find work assigned to you and execute it.

Do not use it for throwaway local TODOs or for code-only tasks that don't belong in a PPM tool.

First moves (every session)

1. agent_bootstrap — one call returns your identity + org, what you may do on this plan/role, each tool's fields, and your assigned-task count. Call this first. (whoami + describe_capabilities + describe_schema individually only if you need more detail.)
2. To find a project or entity by name when you lack an id: search or search_org_knowledge (full-text). Most write tools need an id, not a name.

Availability scales with the org's plan tier and the caller's role; a tool you can't use returns a clear plan/permission error rather than failing silently. When a write is denied, surface the reason (upgrade plan / request role / not a project member) instead of retrying blindly.

Tool families (the map)

• Discovery: agent_bootstrap, whoami, describe_capabilities, describe_schema, search, search_org_knowledge, list_projects, get_project, lookup_help_topic.
• Projects: create_project, update_project, get_project_history, find_similar_projects, summarize_project.
• Tasks: list_tasks, get_task, create_task, add_subtasks (up to 25 children atomically), bulk_create_tasks, update_task, bulk_update_tasks, bulk_update_status, assign_task, bulk_assign, move_task_to_epic, move_task_to_sprint, list_assigned_tasks, list_my_tasks, list_overdue, list_related_tasks.
• Dependencies: link_dependency, unlink_dependency, list_dependencies (FS/SS/FF/SF + lag; cycles are rejected).
• Agile: create_sprint_with_tasks, list_sprints, get_sprint, update_sprint, create_epic, list_epics, get_epic, update_epic.
• Milestones: create_milestone, list_milestones, get_milestone, update_milestone.
• Risks & issues: analyze_project_risks (AI), list_risks, dismiss_risk, create_issue, list_issues, update_issue, resolve_issue, link_issue.
• Governance (Enterprise): , , , , , , .

Call describe_capabilities for the full, plan-resolved list and describe_schema for an entity's exact fields before composing a write.

Common workflows

1. Stand up a project from a brief

1. create_project with a clear name, description, and (optionally) start/end dates.
2. Decompose: prefer bulk_create_tasks for a flat list, or create_task + add_subtasks for a parent with up to 25 children in one atomic call. Write task titles a non-technical PM can read.
3. Group work: create_epic for themes, create_sprint_with_tasks to plan an iteration, create_milestone for delivery dates.
4. Wire order with link_dependency (FS/SS/FF/SF, optional lag).
5. Assign with assign_task / bulk_assign. Confirm with get_project or summarize_project.

2. Check status & Earned Value

• Quick read: summarize_project (progress, open risks/issues, top items) or generate_status_report.
• Overdue: list_overdue. Velocity: get_project_velocity.
• EVM: detect_evm_data_source first (imported costs vs. rate cards vs. none), then calculate_evm for PV/EV/AC/SPI/CPI, compose_evm_narrative for a plain-English summary. Capture a capture_baseline before tracking variance.

3. Risk & issue analysis

• analyze_project_risks runs AI detection and returns findings; persist what matters with create_issue (a materialized problem) vs. leaving it as a risk (a probabilistic future). list_risks / list_issues to review, resolve_issue to close with evidence.

4. Log and review time

• create_timesheet_entry (taskId + hours + date), then submit_timesheet_week. Check week_status / weekly_compliance. Approvers use approve_timesheet_week.

5. Execute assigned work (agent loop)

• list_assigned_tasks → read_agent_brief(taskId) for context → assign_task to claim → update_task status IN_PROGRESS → do the work, posting append_log / create_comment → update_task DONE with evidence (or BLOCKED with a reason). File an create_issue if something is wrong.

Conventions

• Dates are calendar-day YYYY-MM-DD for due/start dates; timestamps are full ISO.
• Decompose with add_subtasks (atomic, up to 25) over many single create_task calls.
• One project at a time: resolve the project id first (via search/list_projects), then operate on it.
• Attribute honestly: report DONE only when verified; use BLOCKED with a reason otherwise.
• Mixed-currency portfolios: report per-currency; EVM math is per-task and source-aware.

Example prompts this skill should handle

• "Create a project in Onplana for the website redesign and break this spec into tasks and a 2-week sprint."
• "What's overdue and at risk on the Atlas migration project? Summarize and file the top issue."
• "Show me EVM (SPI/CPI) for the Q3 portfolio and write a one-paragraph status digest for leadership."
• "Log 3 hours today to the 'Data migration dry run' task and submit my week."
• "Find the tasks assigned to me and start the highest-priority one."
• "Set up an epic and milestones for the Project Online → Onplana migration and wire the dependencies."

When you finish a unit of work, reflect it in Onplana (update the task, add a comment) so the project tool stays the source of truth.