Insights connector

Question your nPlan Insights forecasts, risks and schedules from Claude and other AI assistants that support remote MCP connectors.

What you can do

Forecasts: when a project's target milestone is likely to finish (P10 to P90), how far that is from plan, and how the forecast has moved across schedule updates.
Risks and opportunities: the insights nPlan found in a schedule, with their delay causes, affected activities and mitigation actions, plus your risk register and tracked themes.
Schedules: search activities and milestones, look at one activity's dates, float and forecast, follow the driving path to a milestone, see the forecast for every key milestone, and check schedule integrity.
Outcomes: what the forecast means in money and other units (revenue on completion, delay damages, running costs) at P10, P50 and P90, how uncertain that is, and which activities and milestones the values are mapped to. Add, change or remove those values, have AI suggest them from your project documents, and try what-if changes without saving them.
Summaries: read the written project outlook and delay summaries.
Review and act: update a risk's status, impact or owner, add and update mitigation actions, track themes, and create what-if mitigation scenarios.
Set up analysis: create projects, upload schedules and documents, import a schedule from Schedule Studio, create named revisions with scenario overrides, create views, and start a forecast and risk analysis.

Tools that create or change data are marked as such, so your AI assistant can ask before using them. Outcome mappings are the only thing the connector can delete. Changes are made as you, with your Insights permissions, exactly as if you made them in the web app.

Requirements

An nPlan Insights account. Learn about Insights.
The assistant sees exactly the projects you can see in the Insights web app.

Connect

Connector URL:

https://mcp.insights.nplan.io/mcp

Add the connector in your AI assistant. In Claude, add a custom connector with the URL above. Leave any client ID and client secret fields empty.
You are sent to the nPlan sign-in page. Sign in as you normally do for Insights.
Review the consent page and approve the connection.

Example prompts

Which of my Insights projects are forecast to finish late?
What's the P50 and P80 forecast for completion on the Northern Line project, and how has it moved over the last six updates?
What are the top open risks on this project, and what's causing them?
Show me the driving path to the target milestone.
How healthy is the latest schedule? What's pulling the rating down?
Summarise the project outlook.
Dismiss the scaffolding risk as not a risk, and add an action for Sam to resequence the steel deliveries by 1 November.
What are the P50 delay damages on this project, and how have they changed since the last update?
What would the forecast be if we mitigated the top three risks?

Tools

Read only

List projects `insights_list_projects`	Answers 'which projects can I see' or 'what is the uid of the Thameslink project'. List the Insights projects you can access: uid, name, client and planned completion date. Archived projects are left out unless include_archived is true.
Get project `insights_get_project`	Answers 'what views and key milestones does this project have'. Project details, key milestones and its views (one per analysed target milestone). Use a view's name with the other tools.
Get view `insights_get_view`	Answers 'is this view analysed yet', 'what is its target milestone' or 'how many insights are still to review'. A view's target milestone, latest revision, analysis status or error, and insight review counts.
List revisions `insights_list_revisions`	Answers 'which uploads does this view have' or 'what is the latest data date'. Schedule revisions (uploads) analysed in a view, newest data date first: name, data date, type and processing stage.
Get forecast `insights_get_forecast`	Answers 'when will we finish', 'how late are we against plan' or 'how likely are we to hit the date'. Forecast for the view's target milestone in one revision: planned finish, P10/P20/P50/P80/P90 forecast finish, slip against plan in days, likelihood of meeting the planned date and productivity uplift needed.
Get forecast trend `insights_get_forecast_trend`	Answers 'is the forecast getting better or worse over time'. Use this instead of calling insights_get_forecast once per revision. How the target milestone's planned and P50/P80 forecast finish moved across the view's revisions, oldest first.
Search insights `insights_search_insights`	Answers 'what are the biggest risks', 'which opportunities are still open' or 'what is due this month'; with group_by, 'which WBS branch, calendar or theme has the most risk'. Risks and opportunities nPlan found in a view, paged: name, title, category, status, impact, owner and due date. Filter by status, type and due date range. Use insights_get_insight for detail. For exports, all true returns every row (up to 5,000).
Get insight `insights_get_insight`	Answers 'why is this a risk', 'which activities does it affect' or 'what are we doing about it'. One insight: description, delay causes, affected activities (first 25), actions and matched risk register entries.
Get risk register `insights_get_risk_register`	Answers 'what is in the client's risk register for this view'. Entries from the risk register selected for a view, paged. With group_by, totals per WBS branch, calendar or theme of their activities. For exports, all true returns every row (up to 5,000).
List themes `insights_list_themes`	Answers 'which groups of work are driving delay'. Use this instead of grouping insights_search_tasks results yourself. Tracked themes (groups of related activities driving delay) in a view: name, description, delay sensitivity and activity count.
List actions `insights_list_actions`	Answers 'what mitigations are planned', 'who owns them' or 'which are overdue'. Mitigation actions recorded against a view's insights and themes, paged: title, status, assignee, due date and parent insight or theme. For exports, all true returns every row (up to 5,000).
Search activities `insights_search_tasks`	Answers 'find activity X', 'what are the dates and float of the piling activities' or 'list the activities in this revision'; with group_by, 'which WBS branch, calendar, theme or insight carries the most delay'. Search a revision's activities and milestones by code or name, paged: planned and actual dates, float and P50/P80 forecast finish. Pass page_token from the previous result to get the next page. For exports, all true returns every row (up to 5,000).
Get activity `insights_get_task`	Answers 'when will this activity finish', 'how likely is it to finish on time' or 'what drives it'. One activity or milestone by code: planned, actual and remaining dates, float, forecast start and finish percentiles, likelihood of meeting its planned finish, and predecessors and successors (first 25).
Get driving path `insights_get_driving_path`	Answers 'what drives the end date' or 'which activities are most critical'. Use this instead of walking predecessors with insights_get_task. The dominant driving paths to the view's target milestone (or another activity), in order, with dates, float and criticality; plus the most critical activities overall. Not the whole network.
Get schedule health `insights_get_schedule_health`	Answers 'is this schedule fit to forecast' or 'what is wrong with its logic'. Schedule integrity for a revision: overall RAG rating and reasons, and every integrity check with its count or value and rating, worst first.
Get milestone forecasts `insights_get_milestone_forecasts`	Answers 'when will each milestone finish' or 'which milestones are slipping'. Use this instead of looking milestones up one by one with insights_get_task. Forecast finish of a revision's milestones: planned finish, P10-P90 forecast finishes and slip, likelihood to meet the planned date and whether it is complete. scope key (default) lists the project's key milestones and the view's target; if the project has none, all milestones are listed.
Get view summary `insights_get_view_summary`	Answers 'give me the headline for this project' or 'where are we losing time'. A stored written summary of a view (markdown, truncated): project outlook, current areas of delay or schedule quality trend.
Get job status `insights_get_job_status`	Answers 'is the analysis or forecast finished yet' and returns its result once it is. Wait for background work and return its result. With revision_name, waits for that revision to be forecast. Otherwise waits for a view's analysis job (job, default the newest) and lists recent jobs. With outcome_inference, waits for a revision's outcome mapping inference and counts the drafts it left. Once finished, result has the milestone's new P50 and P80 finish and the change in days from the parent revision.
Get outcomes `insights_get_outcomes`	Answers 'how much money is at risk', 'what will delay cost us' or 'what revenue does this forecast earn'. Use this instead of summing outcome mappings yourself. Outcomes (earned value): the value a revision's forecast earns or costs, from the values mapped to its activities and milestones (revenue on completion, liquidated damages per day of delay, running costs, ...). Per unit and value type: the final cumulative value at P10, P50 and P90, and the uncertainty (P10 to P90 spread as a share of P50) with the rating the Outcomes page shows. include_timeline adds the cumulative P10/P50/P90 curve, one point per month. compare_revision_name adds the P50 change against another revision.
List outcome mappings `insights_list_outcome_mappings`	Answers 'which values are mapped to which activities' or 'which AI suggestions still need review'. The values mapped to a revision's activities and milestones for Outcomes: task code and name, value type, quantity, unit, label and the document it came from. status draft lists AI-suggested mappings still awaiting review; confirmed (default) lists the ones used in the outcomes; all lists both. For exports, all true returns every row (up to 5,000).
List outcome units `insights_list_outcome_units`	Answers 'which units can I map values in'. Units outcome values can be mapped in (id, code, label, dimension), e.g. USD in currency or MWh in energy. Units are shared across all projects.
List documents `insights_list_documents`	Answers 'which documents are uploaded to this project'. Documents uploaded to a project (contracts, specifications, reports): name, display name, type and processing status. Pass names to insights_infer_outcome_mappings.
Get Barry conversation `insights_get_barry_conversation`	Use to read a Barry conversation or fetch an answer that was still being written. Messages in a Barry conversation, oldest first, paged. Waits for an answer Barry is still writing by default, so use it to pick up an insights_ask_barry answer that was still running.
List Barry conversations `insights_list_barry_conversations`	Answers 'what have I asked Barry about this project before'. Your Barry conversations on a project, newest first: conversation_id, title, start time and link.
List Barry skills `insights_list_skills`	Answers 'which saved Barry workflows can I run on this project'. Barry skills (saved multi-step prompts) for a project: the project's favourites, suggestions for it, then the shared library. Each has its steps and the inputs insights_run_skill needs (label and type: MILESTONE, ACTIVITY, FILE, TEXT...).
List saved widgets `insights_list_widgets`	Answers 'which widgets have I or my team saved'. Widgets in the user's My Widgets and widgets shared with them, newest first, optionally for one project: description, owner, the prompt that made it and its conversation. They only render in Insights.

Create or change data

Compare two revisions `insights_compare_revisions`	Answers 'what changed between these two revisions and why' or 'which activities lost the most float since last month'. Compares revision A with revision B, milestone by milestone and activity by activity (paired by activity code, as task IDs change between uploads). For each: planned finish, P10-P90 forecast finish, planned float, float to the target milestone (from the links' free float), the forecaster's deterministic float, and float at P10-P90 in A and B, with the change B minus A in days. Float at P80 is the pessimistic end: 80% of simulations have at least that much. Float is measured to milestone_code (default the view's target), and is worked out by a backward pass the first time a revision and milestone are asked for. scope: all (default; activities in both revisions, biggest P80 change first, plus counts of added and removed ones), critical (the critical path to the milestone and up to 5 near-critical paths) or driving (up to 5 driving paths). narrative is Schedule Comparison's executive summary of what changed between the two schedules; the tool waits for it by default, and calling again returns it once ready. Starts a comparison in Schedule Comparison, which the user can also open there.
Update insight `insights_update_insight`	Use to accept, dismiss or assign a risk or opportunity. Review a risk or opportunity: change its status, impact, owner, due date or notes. Dismiss it by setting status CLOSED_DISMISSED with a dismiss_reason. Only the fields you pass change.
Add action `insights_add_action`	Use to record a mitigation against an insight or theme. Add a mitigation action to an insight or to a tracked theme. Pass exactly one of insight_name or theme_name.
Update action `insights_update_action`	Use to reassign, reschedule or close an action. Change an action's title, detail, assignee, due date or status. Only the fields you pass change.
Track theme `insights_track_theme`	Use to follow a theme across future revisions. Start tracking a theme so it appears in the view's tracked themes.
Untrack theme `insights_untrack_theme`	Use to stop following a theme. Stop tracking a theme. Its actions are kept.
Create view `insights_create_view`	Use to forecast a different target milestone. Create a view for a target milestone of a project, or update the view that already targets it. Optionally attach a ready revision; attaching starts no new run. Run insights_run_analysis afterwards only if the revision's risks still need finding, which can take hours.
Run analysis `insights_run_analysis`	Use to forecast and find risks in a revision that has not been analysed. Run nPlan's forecast and risk analysis of a view's revision (default the latest). Takes from under an hour to many hours: time grows with the threat and opportunity counts and the schedule's size, and hundreds on a large schedule can take over 10 hours. Until it finishes, the website blocks new schedule uploads to the project. Waits for it by default and returns the target milestone's P50 and P80 and the change from the parent revision. Threat and opportunity counts default to the view's last successful analysis, else 25 each; reusing more than 25 is refused until the counts are passed explicitly, so ask the user which size to run first.
Create mitigation scenario `insights_create_mitigation_scenario`	Answers 'what if we mitigate these risks: how much time do we win back'. Create a what-if scenario from a view's latest revision in which the given insights are mitigated. Effectiveness is the fraction of each insight's delay removed (default 0.6). The scenario is a new revision; waits for its forecast by default and returns the target milestone's P50 and P80 and the change from the revision it came from.
Create project `insights_create_project`	Use to start a new project before uploading its schedule. Create an empty Insights project. client_id is needed only if you belong to more than one client; an error lists them.
Create upload URL `insights_create_upload_url`	Use to upload a file too large to send inline. Get a one-off signed URL to upload a large or binary file (up to 1 GB for a schedule, up to 30 MB for a document) to. PUT the raw file bytes to upload_url with each of upload_headers sent unchanged. Copy each header value from upload_headers, because a schedule URL and a document URL have different values (e.g. curl -T <file> -H 'x-goog-content-length-range: <value from upload_headers>' '<upload_url>'). Then pass upload_id to insights_upload_schedule or insights_upload_document. Cloud Storage rejects the PUT if a header is missing or changed, or if the file is over the signed limit (upload_max_bytes). For a document, ingest_max_bytes is the same as upload_max_bytes, so insights_upload_document can ingest every document that uploads. Files under 10 MB can be passed to those tools directly instead.
Upload schedule `insights_upload_schedule`	Use to add a new schedule file to a project. Upload a schedule file (.xer, .mpp, .xml) to a project. Insights parses it and creates an unprocessed (LIGHT) revision; waits for that by default. Next, create a forecastable revision from it with insights_create_revision.
Import from Schedule Studio `insights_import_from_schedule_studio`	Use to forecast a schedule built in Schedule Studio. Import a Schedule Studio schedule into an Insights project as a new unprocessed (LIGHT) revision. The schedule is copied as is, with no XER or other file conversion. Waits for the revision by default. Next, create a forecastable revision from it with insights_create_revision.
Upload document `insights_upload_document`	Use to give Barry and risk matching a contract, report or risk register. Upload a document (at most 30 MB) to a project, for Barry and risk matching. It is processed in the background.
Create revision `insights_create_revision`	Answers 'what if this activity took longer or the data date moved' by forecasting a variant of a revision. Create a named revision from an existing one (e.g. a just-uploaded schedule), with scenario overrides. Waits for its forecast by default and returns the milestone's P50 and P80 and the change from the parent revision. To analyse its risks, attach it to a view with insights_create_view, then run insights_run_analysis.
Ask Barry `insights_ask_barry`	Answers 'what do the contract documents say about access dates' and other open questions the other tools don't cover. Ask Barry, nPlan's schedule assistant, a question about a project, or continue a conversation with conversation_id. Barry can query the schedule and forecasts, search the project's documents and build widgets. Waits for the answer by default and returns it as markdown, with the widgets in it described (they render in Insights, at link). When the question is about a particular revision or scenario, pass revision_name when starting the conversation so Barry answers from it rather than asking which revision to use.
Run Barry skill `insights_run_skill`	Use to run a saved multi-step Barry workflow, such as a monthly report. Run a Barry skill in a new conversation on project_uid, or in conversation_id, and wait for it. Returns Barry's answer to each step, the last in full. Ask the user for any inputs you can't find with the other tools (e.g. a milestone code from insights_get_milestone_forecasts).
Create Barry widget `insights_create_widget`	Use to make a chart or table for a dashboard or report. Ask Barry to build a widget (a chart, table or dashboard tile from the project's data) and save it to the user's My Widgets in Insights. Returns widget_id, Barry's answer and a link; the widget itself only renders in Insights.
Save Barry widget `insights_save_widget`	Use to keep a widget Barry already built in a conversation. Save a widget from a Barry answer to the user's My Widgets. Defaults to the first widget in the latest answer that has one.
Set outcome mappings `insights_set_outcome_mappings`	Use to say what an activity or milestone is worth, e.g. damages per day of delay or a payment on completion. Add or change the values mapped to a revision's activities and milestones for Outcomes. Each entry with an id (from insights_list_outcome_mappings) changes only the fields passed; each entry without one adds a mapping and needs task_code, value_type, quantity and unit_code (from insights_list_outcome_units). element_type defaults from the task, label to the task name. Value types: FIXED_ON_COMPLETION (once, when the task finishes), PER_DAY_OF_DELAY (per day the task finishes after its planned date, e.g. liquidated damages), POST_COMPLETION_RATE (per day from completion, e.g. revenue once open), RUNNING_COST (per day while the task runs), RUNNING_DELAY_COST (per day the task runs past its planned finish). Entries are applied one by one and failures are reported per entry. Then the outcomes are recomputed once and the new P10/P50/P90 values returned (recompute false skips that).
Delete outcome mappings (destructive) `insights_delete_outcome_mappings`	Use to remove wrong or unwanted outcome mappings. Delete outcome mappings (confirmed or draft) from a revision by id, from insights_list_outcome_mappings, then recompute the outcomes once (recompute false skips that). This cannot be undone.
Copy outcome mappings `insights_copy_outcome_mappings`	Answers 'carry my outcome mappings over to the new revision'. Copies the confirmed outcome mappings of an earlier revision of the same project (source_revision_name, from insights_list_revisions) to this one, matched by activity code, then recomputes the outcomes once (recompute false skips that). Activities missing from this revision and mappings it already has are skipped, so it is safe to run again. New uploads normally get their mappings copied automatically; use this when they did not, instead of re-creating them with insights_set_outcome_mappings.
Create outcome unit `insights_create_outcome_unit`	Use only when no existing unit fits a value. Add a unit outcome values can be mapped in, e.g. code tCO2e, label Tonnes of CO2 equivalent, dimension carbon. Units are shared by every project and customer, so check insights_list_outcome_units first; an existing code is refused.
Compute outcomes `insights_compute_outcomes`	Use after changing mappings to refresh the outcome values. Recompute a revision's outcomes from its confirmed mappings and the latest forecast, and return the new P10/P50/P90 values. reporting_horizon caps per-day accruals (default: the latest P100 finish plus two years).
Run outcome scenario `insights_run_outcome_scenario`	Answers 'what if the damages rate or a payment changed' without saving anything. What-if outcomes, not saved: the revision's confirmed mappings with changes applied, computed against its forecast. A change with a mapping_id overrides that mapping's task_code, value_type, quantity or unit_code, or drops it with remove true; a change without one adds a mapping (task_code, value_type, quantity and unit_code). Returns P10/P50/P90 and the P50 change against the saved outcomes.
Infer outcome mappings `insights_infer_outcome_mappings`	Use to have AI read contracts for values to map, instead of mapping them by hand. Start AI inference of outcome mappings from project documents (from insights_list_documents): it reads them for values such as delay damages, milestone payments and running costs and maps them to the revision's activities and milestones as drafts. Nothing is confirmed until you call insights_confirm_outcome_drafts. It usually takes 40 to 60 minutes, so expect to follow it with insights_get_job_status (outcome_inference true).
Confirm outcome drafts `insights_confirm_outcome_drafts`	Use to accept AI-suggested outcome mappings so they count. Confirm AI-suggested outcome mappings (from insights_list_outcome_mappings status draft) so they count in the outcomes: draft_ids, or all true for every draft on the revision. Then recompute once and return the new P10/P50/P90 values (recompute false skips that).

Data and security

You sign in through nPlan's own sign-in page. Your password never reaches the AI assistant, and the assistant is never given your nPlan session.
The assistant gets a connector-specific access token that lasts one hour and is refreshed automatically. Refresh tokens are single-use. A connection that isn't used for 30 days expires, and every connection must be re-approved after 90 days.
The connector does not store your projects or schedules. It passes each request to Insights with your permissions and returns the result. It stores only what it needs to keep you connected: hashes of the tokens it issues, and your nPlan sign-in refresh token encrypted with Google Cloud KMS. Tokens are never logged.
Results returned to the assistant are processed by its provider under their terms. See our privacy policy for how nPlan handles your data.

Disconnect

Remove the connector in your AI assistant's settings. Its tokens then stop being used and expire. To have your connection revoked immediately, email support@nplan.io.

Support

Email support@nplan.io with the time of the problem and the assistant you were using.