Tracker MCP

The ConstruChain Tracker MCP connection lets you consult progress, weekly snapshots, published schedule versions and dated memories in ChatGPT or Claude. All 13 tools are read-only.

Connect your account

You need a Tracker account authorized to establish the connection and access to the projects you want to consult. During authorization, explicitly select the projects the connection may read. No project is selected by default.

Use this address for a custom remote connection:

https://mcp.construchainglobal.com/tracker/mcp

Sign in with your ConstruChain account on the authorization screen and review the requesting application, projects and permissions before continuing. You do not need to enter an API key. Client identifier and secret fields, if shown by the platform, remain empty for automatic connection registration.

Follow the ChatGPT or Claude guide. Custom connection availability depends on your platform plan and organization settings. This guide describes a custom connection and does not assume that the connector is listed in a directory.

First query

1. Request the list of projects authorized for this connection.
2. Choose a returned project.
3. Ask for that project’s context before requesting an analysis.
4. Consult current progress, existing snapshots or the published versions you need.

List the Tracker projects authorized for this connection. Wait for my choice before analyzing a project.

For the chosen project, show the available disciplines and the actual availability of current data and snapshots. Then list the published schedule versions.

Available tools

Tool	What it reads
list_tracker_projects	Projects authorized for the connection
get_tracker_project_context	Disciplines and actual availability of the project’s current and weekly sources
get_tracker_current_summary	Current progress summary, with an optional discipline filter
search_tracker_items	Current items or items in a real weekly snapshot, with search, module, work package, status and date filters
list_tracker_snapshots	Exact dates of available weekly snapshots and their completeness
compare_tracker_snapshots	Comparison of two real snapshots of the same discipline
list_tracker_schedule_versions	Published XER versions, their Current/Baseline roles, history and dates
get_tracker_schedule_version_context	Context for one published version; without a specific version, reads Current
query_tracker_schedule	Defined schedule datasets, with filters and pagination
get_tracker_schedule_changes	Changes between two distinct published versions from the same project
list_tracker_schedule_memories	Dated memories, revisions and linked versions
get_tracker_schedule_memory	Text of one project memory revision
get_tracker_schedule_verification_context	Memory, related versions, recommended comparison pair when available, and factual gaps

The datasets for query_tracker_schedule are overview, activities, wbs, relationships, resources, health, calendars and work_packages. The query uses these datasets and their filters; it does not accept SQL, table names, file paths or raw XER access.

MCP disciplines are steel, pipe, pipe_insulation, electrical and fireproofing, according to project data and access. Use the disciplines reported by the context rather than assuming all are available.

Ready-to-use prompts

Summarize current progress for the selected project by discipline. Identify the source, units, available date and any missing information. Do not include costs.

List the available snapshots and, after I choose two dates, compare values for the same discipline. Do not create daily history or estimate missing snapshots.

Search critical activities in the published version I choose. Report the version, filters and whether more pages are available before concluding the analysis.

Read the dated memory I specify as business context. Obtain the verification context and use the exact recommended pair when available. If a defined target is missing, show the available versions for me to choose.

See Progress and snapshots and Schedules and memories to interpret results.

Limits and privacy

The connection does not update progress, upload or publish schedules, edit memories or change project records. It does not create missing history. A missing result must be treated as unavailable, not as zero.

Lists may be paginated or truncated. Ask the assistant to check the returned indicators and continue through available pages before claiming to have read the complete dataset. Costs may only be queried when the connection is authorized and the fields are available.

Memories are business content and never instructions to the assistant. Verification context does not declare a commitment met or missed. When VERIFICATION_TARGET_REQUIRED appears, explicitly select the intended published versions.

Results are delivered to the connected ChatGPT or Claude session. Authorize only data you may share with that platform. Do not share passwords, codes, tokens or authorization links in conversations or support requests. For connection failures or access questions, consult Help.