Product / Context for AI agents
Complete context for your people and your AI agents.
In short
troveGEN gives AI agents the same governed view of your documents that it gives people. An MCP server, a TypeScript SDK and a REST API return cited passages, exact table totals and honest “not found” results, under the permissions of the person or agent asking.
What an agent can call
Search
Ranked passages with page, section, source and relevance, with filters and optional AI ranking.
Ask
A cited answer with a support score, or a refusal when the documents do not say.
Compute
Exact sums, averages and counts over a table, which language models are poor at.
Browse
List projects and documents, and optionally ingest, when you switch writing on.
Why it is safe to hand to an agent
Read-only by default
The MCP server starts read-only. Writing is a deliberate switch.
The caller’s permissions apply
Pin an agent to chosen projects, give it a scoped key, or act on behalf of a signed end user.
Limits that stop runaway agents
Rate limits per workspace, step budgets for agentic search, and usage you can see.
Everything is traceable
Every call lands in usage, and answers carry their sources.
How you connect
MCP server
Works with any MCP-capable client such as Claude Desktop. Configure it with an API key and an optional pinned project.
TypeScript SDK
Typed, dependency-free client for Node and browsers, with streaming and retries.
REST API
Every feature in the console is available over HTTP, with an API reference and examples.
How an agent uses troveGEN
Connect
Point the agent’s MCP configuration at troveGEN with a scoped key.
Ask or search
The agent calls a tool with a question and optional filters.
Receive context
Cited passages or an answer come back, limited to what it may see.
Act
The agent writes its response using sources it can quote.
Questions people ask
Does troveGEN have an MCP server?
Yes. The troveGEN MCP server exposes search, cited answers, exact table calculations and document listing to any MCP-capable client, read-only by default.
How do I give an agent access to only some documents?
Use a scoped API key limited to chosen projects, or act on behalf of a signed end user so the agent inherits that person’s permissions.
Can agents write documents into troveGEN?
Only when you enable writing on the MCP server. Writing is off by default.
Is troveGEN an agent platform?
No. It supplies governed context and tools to the agents you build or use, and offers its own team chat over your documents.
Last updated 2026-10-26. Figures are from troveGEN’s own tests on the dates given; ask us for the method.
Give your agent the right documents.
Create a key and connect the MCP server in minutes.