“Show me invoices awaiting approval that contain an anomaly.”
The assistant searches documents by status, company, customer or period, then summarises the points to review.
Tempolia AI and MCP server
Make the most of your data with Tempolia AI and our MCP server
Tempolia AI, included in all our plans, helps you make the most of your management data. You can also centralise your agentic capabilities and use our MCP server alongside MCP servers from other software. With 75 structured actions, Tempolia's MCP server enables an authorised agentic AI to search, analyse or update objects within 144 permission levels.
75structured MCP actions to search, analyse and act
144Tempolia permission levels applied to every call
Agentic AIchain tools with approval and traceability
Compatible with any AIthrough MCP or an authorised tool gateway
Open ecosystem
Choose your AI. Tempolia remains in control.
Claude, Gemini, Mistral, Perplexity, GitHub Copilot, DeepSeek or another assistant can be part of the solution when its environment can use MCP or an authorised tool gateway.
Claude
Gemini
Mistral AI
Perplexity
GitHub Copilot
DeepSeekPractical examples
Natural-language requests connected to Tempolia objects.
“Which engagements are over their time budget this month?”
It brings together the budget, actuals, invoicing and remaining work within the scope available to the user.
“List customers whose invoices have been overdue for more than 30 days.”
It uses balances, due dates and payments to prepare a reminder review.
“Add 1 hour 30 minutes for a meeting on the ACME engagement after I confirm.”
The MCP server can call the time-creation tool with the same validations and permissions as the interface.
“Prepare a 360-degree customer summary before my meeting.”
The response can bring together engagements, budgets, invoices, balances, events, documents and the latest authorised activities.
“Which staff members have not completed their time entry for the week?”
The assistant identifies incomplete periods according to the manager's rules and scope.
“Create the 2026 engagement for these customers from the approved template.”
A batch action can be prepared, presented for confirmation and executed with traceability.
“Export revenue and recovery by activity.”
Tempolia AI uses authorised reports and returns the results as a summary or file.
From text to action
A controlled four-step process.
Understand
The assistant identifies the object, filters, period and expected result.
Authorise
Tempolia checks the user, permissions and data scope.
Confirm
Creation or modification actions may require explicit confirmation.
Record
The call, result and sensitive operations are logged according to their nature.
MCP
A standard protocol for connecting assistants to Tempolia tools.
The MCP server exposes 75 structured actions for search, retrieval, analysis, creation or updates across customers, engagements, time, expenses, tasks, staff, documents and reports.
The protocol does not provide unrestricted database access. Every call passes through Tempolia authentication and permissions. The available tools can be restricted by profile.
Targeted read accessSearch and retrieve authorised objects.
Controlled actionsCreate or edit after validating parameters and permissions.
Structured responsesData that the assistant can use without direct SQL access.
TraceabilityLog calls and sensitive operations.
Necessary safeguards
AI cannot bypass business rules.
The same scope as the user
The assistant cannot see data that the signed-in profile cannot access.
Confirmation before sensitive actions
A creation, modification or deletion may require explicit approval.
No silent modifications
Approved objects, particularly tax-related records, remain subject to integrity and correction rules.
Verifiable results
The filters, sources and objects used must be explainable to the user.
Tempolia AI / MCP FAQ
Frequently asked questions before activation.
Does the assistant access the database directly?
No. The MCP server exposes structured tools: the assistant calls a Tempolia action with controlled parameters and has no unrestricted SQL access. Every call is authenticated and filtered by the user’s permissions and data scope; the result and sensitive operations are logged according to their nature.
Can it modify data?
Only when the relevant creation or update tool is enabled and the user has the required permissions for the object and scope concerned. Tempolia applies the same validations as the interface; explicit confirmation may be required before the action, which remains traceable.
Which assistants can use the MCP server?
Claude, Gemini, Mistral, Perplexity, GitHub Copilot, DeepSeek or any assistant whose environment supports MCP or an authorised tool gateway can be connected using the appropriate Tempolia settings. Compatibility grants no extra permissions: every call remains subject to authentication, profile and data scope.
Can AI be restricted to read-only access?
Yes. The tool scope can be limited to search, retrieval, analysis and reporting, without exposing creation or update actions. Tempolia permissions also continue to filter visible companies, objects and data; write tools are enabled only for a defined and tested use case.
How should an initial use case be defined?
Choose a frequent, measurable question, then specify the data, filters, permissions, possible actions and expected result. Test the case in a controlled environment with explicit success criteria; sources, objects used, confirmation steps and responsibilities must remain understandable to the user.
What do the 75 MCP actions cover?
They cover search, retrieval, analysis and, subject to permissions, the creation or update of customers, engagements, time, expenses, tasks, staff, documents and reports. The actions actually available depend on profile and enabled scope; they return structured responses without opening direct database access.
How do the 144 permission levels apply to AI?
Every tool call is authenticated and checked by Tempolia. The assistant inherits the authorised scope for companies, features, objects, actions and data and cannot widen it; interface validations still apply, while confirmation and logging can be required for sensitive operations.
Begin with a measurable use case.
We define the question, data, permissions, actions and success criteria.