Iterative research with configurable depth and breadth for complex topics.
Maintain this server? Claim it
Call this server's tools from your browser. Nothing to install.
Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.
Tool list from our last check. Connect live to refresh it.
Conduct deep, iterative research on a topic by generating multiple search queries, processing the results, and recursively exploring new research directions.
The main topic or question to research
The specific goal or objective of the research
Depth of recursive exploration (0-2)
Number of search queries per research direction (1-5)
Optional connection ID for the user
Maximum number of search queries to process (2-5)
Maximum duration for the research process in seconds (60-300)
This tool isn't marked read-only, so it may change something on the server or in an account you connected.
Pick your client. Values in capitals, like YOUR_TOKEN, are yours to fill in.
claude mcp add --transport http mcp-deep-research https://nova.civai.co/mcp/agents/mcp-deep-research
Conduct deep, iterative research on a topic by generating multiple search queries, processing the results, and recursively exploring new research directions.
| Argument | Type | Description |
|---|---|---|
| depth | integer | Depth of recursive exploration (0-2) |
| topic* | string | The main topic or question to research |
| breadth | integer | Number of search queries per research direction (1-5) |
| objective* | string | The specific goal or objective of the research |
| connection_id | string | Optional connection ID for the user |
| max_total_queries | integer | Maximum number of search queries to process (2-5) |
| max_duration_seconds | integer | Maximum duration for the research process in seconds (60-300) |
{
"type": "object",
"required": [
"topic",
"objective"
],
"properties": {
"depth": {
"type": "integer",
"default": 1,
"maximum": 2,
"minimum": 0,
"description": "Depth of recursive exploration (0-2)"
},
"topic": {
"type": "string",
"description": "The main topic or question to research"
},
"breadth": {
"type": "integer",
"default": 3,
"maximum": 5,
"minimum": 1,
"description": "Number of search queries per research direction (1-5)"
},
"objective": {
"type": "string",
"description": "The specific goal or objective of the research"
},
"connection_id": {
"type": "string",
"default": "",
"description": "Optional connection ID for the user"
},
"max_total_queries": {
"type": "integer",
"default": 2,
"maximum": 5,
"minimum": 2,
"description": "Maximum number of search queries to process (2-5)"
},
"max_duration_seconds": {
"type": "integer",
"default": 120,
"maximum": 300,
"minimum": 60,
"description": "Maximum duration for the research process in seconds (60-300)"
}
}
}Generate a comprehensive research report from existing research data.
| Argument | Type | Description |
|---|---|---|
| objective* | string | The specific goal or objective of the report |
| research_data* | string | The research data and findings to include in the report |
{
"type": "object",
"required": [
"research_data",
"objective"
],
"properties": {
"objective": {
"type": "string",
"description": "The specific goal or objective of the report"
},
"research_data": {
"type": "string",
"description": "The research data and findings to include in the report"
}
}
}We read every tool's name, description and inputs the way the model will, looking for instructions hidden from you. A clean result means no warning signs, not a guarantee. Scanned Nov 2, 2026.
No readOnlyHint / destructiveHint annotations, so clients can't tell a read from a write without asking.
Fix: Add annotations to each tool, e.g. { readOnlyHint: true } for tools that only read.