MCP-connected assistants
Expose your connected tools and defined workflows to compatible AI clients through MCP.
THE SECURITY LAYER FOR MCP
Give your team’s AI assistants access to business tools—with precise limits on the records they read, the data they return, and the actions they take. Nectry enforces those limits outside the model.
Exactly the authority the task requires.
01 FINE-GRAINED ACCESS CONTROL
Decide more than whether an agent can use a tool. Define which records it can read, which fields it can return, and who it can send data to. Keep useful access within the boundaries your team sets.
Explore the policy model. Available scopes depend on the integration and deployment; we’ll map your requirements in a demo.
Limit which records and response fields can reach the model.
Only records assigned to the agent’s team
Only the records the task needs
Express the logic between tool calls
Keep inputs and actions within scope
Within the workflow’s authority
02 CUSTOM TOOLS & WORKFLOWS
Turn a workflow across your systems into a purpose-specific MCP tool. Define its inputs, operations, and permissions together, so an agent gets the capability the task needs with clear limits on what it can do.
ILLUSTRATIVE TOOLLog a call outcome.Update the call notes and next step, with other CRM fields outside the tool’s scope.
Explore your workflow with us →03 DETERMINISTIC ENFORCEMENT
Give agents useful access without handing over unrestricted authority. Nectry checks each governed tool call against your policy and blocks out-of-scope actions before execution—even when the model asks for them.
A model’s request is not authorization.
Identity · data · action · context
Explicit rules. Checked before execution.Keep sensitive data in bounds.Limit which resources an agent can reach.
Protect consequential actions.Constrain writes, deletions, and other changes.
Keep authority outside the model.A generated request cannot grant itself permission.
04 FORMAL METHODS, PRACTICAL CONTROL
NectryCore uses a restricted language that makes supported security properties decidable. The compiler can check what a program may access, which tools it may invoke, and where authorization is required.
Formal verification applies to supported properties of compiled NectryCore programs. MCP runtime checks enforce requests on the governed path; they do not prove the entire behavior of an external agent.
What data can reach which output
Which systems may be contacted
Which calls and argument shapes are permitted
Where human authorization is required
Chief Scientist Adam Chlipala
05 FITS THE WAY YOUR TEAM WORKS
Keep the assistant experience familiar. Bring supported tools and workflows through Nectry, with an integration and deployment approach matched to your environment.
Expose your connected tools and defined workflows to compatible AI clients through MCP.
Bring supported integrations together and define the operations your workflows need.
Discuss Docker-based deployment, local-model requirements, and the governance controls your organization needs.
Advanced governance is evolving with enterprise partners. We’ll distinguish available capabilities from planned controls when mapping your requirements.
MAKE ROOM FOR AI. KEEP AUTHORITY WITH YOUR TEAM.
Bring one task your team wants to delegate. See the tools and permissions it would need.
Book a Demo ↗