Company research. For agent tool calls.
A model can describe a company only as well as the context it receives. Design your product around a bounded lookup, a predictable response, and fields your users can inspect.
The same research, through MCP.
- 01Free
get_research_listSaved references + workspace notes - 02Free
get_list_contextPublic context for selected members
A small tool contract. A useful piece of context.
The request below uses the implemented get_company input with a fictional source ID. The illustrative result shows company context and an unknown verification date; it is not a live response.
Start with the question“What could my agent return when a user asks about Forma Labs?”
{
"name": "get_company",
"arguments": {
"id": "fictional-company-forma"
}
}{
"company": "Forma Labs",
"domain": "forma.example",
"industry": "SaaS",
"location": "San Francisco, US",
"employees": "51–200",
"source": "fictional-preview",
"production_verified_at": null
}Design example only. No API request is sent and no live MCP server is connected.
Design the tool around one bounded task.
Specify what the user needs, what the tool accepts, and which fields the answer can cite. Keep unsupported claims outside the generated result.
Describe the job
Give the agent a specific search or lookup task with explicit inputs and result limits.
Return structured context
Design around predictable fields, source information, and clear missing values.
Keep the result inspectable
Show the returned evidence to the person using your product before a consequential next step.
A tool is useful when its boundaries are understandable.
The tool description explains when the capability should and should not be used.
Inputs are validated and search results are bounded.
Missing values stay explicit through the tool response and user interface.
No-match and error states have explicit handling; the schema example does not contact a data service.
Before you get started.
Can I connect Claude, Cursor, Codex, or Hermes today?
Create a verified account at app.vencavo.com and select a workspace. Use the MCP setup guide to connect your client. Client instructions distinguish documented setup from recorded acceptance.
What is available for development now?
Read the implemented search, profile, list, credit, and error contracts. The public examples use fictional records and do not connect to a model or data service.
Why separate a lookup from the rest of an agent workflow?
A narrow tool is easier to describe, validate, and inspect. Your application can decide how to present the result and which actions require additional review.
Make the next conversation
a better-informed one.
Build a company brief with a clear research question, inspectable facts, and an honest list of unknowns.