MCP servers for Vista, Spectrum and ProjectSight
An MCP server that exposes your ERP or project management system to AI agents through a small set of governed tools, authenticated as the user.
The problem
Accounting and project teams want to ask an AI assistant about invoices, commitments, job cost or RFIs, but the ERP's API is large, the data is sensitive, and the agent must only see what the user is allowed to see.
What we build
We build an MCP server around the system's API with tools designed from the user's tasks, a gateway that routes requests to the right operation, on-behalf-of authentication so the agent acts as the signed-in user, and a policy layer that keeps deletes off and gates writes behind approval.
How it works
- 1
Start from the tasks
We list the questions and actions users need and design tools around them, rather than one tool per endpoint.
- 2
Authenticate as the user
The server exchanges the agent platform's token for a product token on behalf of the user, so the system's own permissions apply.
- 3
Apply policy
Reads are allowed, writes need approval or are off, deletes are never exposed by default, and bulk operations are capped.
- 4
Observe
Every tool call is logged with timing and outcome, without storing raw inputs or outputs, so you can see what agents do and where they fail.
- Viewpoint Vista
- Trimble Spectrum
- Trimble ProjectSight
- Model Context Protocol
- Claude
- Microsoft Copilot
The value it creates
Time saved
Answers that take several screens and exports come from one question; we time common lookups before and after.
Visibility
Tool-call telemetry shows which questions users ask and which fail.
Standardization
Every agent uses the same governed tools instead of its own credentials.
Collective intelligence
Knowledge locked in one person's ERP skills becomes available to the team, within their permissions.
Proof
Related use cases
- Built and shown
AI assistant for project drawings, specs and documents
Ask questions of a project's drawings, specs and meeting notes and get answers that cite the file and passage they came from.
- We design and build this
AI review agent for Viewpoint Vista AP unapproved invoices
A read-first agent that works the Vista unapproved invoice queue, scores each invoice for risk, compares it with the commitment, and leaves the approval decision to a person.
- Built and shown
AI sales automation: lead capture, research, follow-ups and proposals
Agents handle the research, scheduling and first drafts around each sales step, a person approves anything customer-facing, and the CRM stays up to date.
- Built and shown
AI schedule builder for CPM schedules with evidence and XER export
Specialized agents draft a CPM schedule from project documents and client answers, show the evidence behind each activity, and export XER for the scheduler to review.
- Built and shown
MCP servers for Procore, Primavera P6 and other construction systems
An MCP server exposes a construction system's API as tools an AI agent can call, with scoped permissions and a log of every call.
Frequently asked questions
Have you built MCP servers for these systems?
Yes, while at Trimble we built MCP servers for ProjectSight, Vista AP invoice review, Accubid, Trimble Connect and Tekla PowerFab, among others. That code belongs to Trimble and is not public; for clients we apply the same patterns to your systems.
Can the agent change data in our ERP?
Only if you decide it should. The default is read-only. Writes can be enabled per domain behind an approval step, and deletes are not exposed unless you explicitly require them.
How does the agent know what the user is allowed to see?
The server exchanges the user's token for a product token, so the ERP applies its own security to every call. The agent never holds a shared admin credential.
Next step
Which report or workflow would you like to improve?
Tell us what your team does today, which systems are involved, and what you want to change. We'll discuss whether there is a practical fit.
Prefer email? charley@buildflows.ai