Curated news search and deep research reports for your reader: tech, AI, business, policy, media.
Maintain this server? Claim it
Call this server's tools from your browser. Nothing to install.
Optional. Your Polymath News API key, sent as a bearer token. Without it, the server answers a few trial searches a day. Create a key at https://www.polymath.news/developers/account.
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.
Search current news for a specific reader. Covers technology, AI, business, policy, and media and entertainment (about 500 curated sources, not all news). Returns one list of stories, best match first: events, each grouped across the outlets covering it, with every article (publisher, date, link), the lead article's excerpt, the names in it, and which of your terms it matches and how well (0 to 1). Then Polymath reports, with title, insight and link. Put topics in "interests": 3 to 8 specific topics, named the way a news desk would ("AI regulation", "data centers"). The name of a person or a company works as an interest too ("Nvidia"). Each interest gets its fair share of the list, and the reply says what each one found, including nothing. "query" is for keywords. "profile" says who you work for: it tailors reports and rarely selects stories by itself. "since" returns only what is new or updated after a time; every reply gives next_since, the value to pass next time. With no arguments it returns the top stories right now. No API key was sent, so this is a trial: 1 searches a day, up to 3 stories per search, report titles and links only.
Keywords or a short question, for example "EU AI Act enforcement" or "who is buying data centers".
Only what is new since this time (ISO 8601, for example "2026-11-09T07:00:00Z"). Pass the next_since value from your previous search_news reply. Overrides since_days. Stories are marked new or updated.
Less or more text. brief (default): every story with every article (publisher, date, link), the lead article's excerpt and the names in it. full: the same, plus every article's excerpt and a summary of each Polymath report.
Who you are working for, in one or two sentences, for example "a mid-size US law firm with a technology practice". The profile tailors reports. It rarely selects stories by itself, so put the topics in "interests".
Three to eight specific topics this reader cares about, which you infer from the profile, for example ["AI regulation", "data centers", "Nvidia"]. Name each topic the way a news desk would. The name of a person or a company works as an interest too. Each interest gets its fair share of the results, and every story says which interests it matches and how well, so specific topics work far better than a profile alone.
How many days back to look. Default 7 for stories and 30 for reports. Stories go back at most 30 days, reports 90.
Most stories and most reports to return, each. Default 8.
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.
Run in a terminal
claude mcp add --transport http polymath-news https://mcp.polymath.news/mcp
Headers
Authorization (optional, not in the config above): Optional. Your Polymath News API key, sent as a bearer token. Without it, the server answers a few trial searches a day. Create a key at https://www.polymath.news/developers/account.Search current news for a specific reader. Covers technology, AI, business, policy, and media and entertainment (about 500 curated sources, not all news). Returns one list of stories, best match first: events, each grouped across the outlets covering it, with every article (publisher, date, link), the lead article's excerpt, the names in it, and which of your terms it matches and how well (0 to 1). Then Polymath reports, with title, insight and link. Put topics in "interests": 3 to 8 specific topics, named the way a news desk would ("AI regulation", "data centers"). The name of a person or a company works as an interest too ("Nvidia"). Each interest gets its fair share of the list, and the reply says what each one found, including nothing. "query" is for keywords. "profile" says who you work for: it tailors reports and rarely selects stories by itself. "since" returns only what is new or updated after a time; every reply gives next_since, the value to pass next time. With no arguments it returns the top stories right now. No API key was sent, so this is a trial: 1 searches a day, up to 3 stories per search, report titles and links only.
| Argument | Type | Description |
|---|---|---|
| query | string | Keywords or a short question, for example "EU AI Act enforcement" or "who is buying data centers". |
| since | string | Only what is new since this time (ISO 8601, for example "2026-11-09T07:00:00Z"). Pass the next_since value from your previous search_news reply. Overrides since_days. Stories are marked new or updated. |
| detail | string | Less or more text. brief (default): every story with every article (publisher, date, link), the lead article's excerpt and the names in it. full: the same, plus every article's excerpt and a summary of each Polymath report. |
| profile | string | Who you are working for, in one or two sentences, for example "a mid-size US law firm with a technology practice". The profile tailors reports. It rarely selects stories by itself, so put the topics in "interests". |
| interests | array | Three to eight specific topics this reader cares about, which you infer from the profile, for example ["AI regulation", "data centers", "Nvidia"]. Name each topic the way a news desk would. The name of a person or a company works as an interest too. Each interest gets its fair share of the results, and every story says which interests it matches and how well, so specific topics work far better than a profile alone. |
| since_days | integer | How many days back to look. Default 7 for stories and 30 for reports. Stories go back at most 30 days, reports 90. |
| max_results | integer | Most stories and most reports to return, each. Default 8. |
{
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {
"query": {
"type": "string",
"maxLength": 300,
"description": "Keywords or a short question, for example \"EU AI Act enforcement\" or \"who is buying data centers\"."
},
"since": {
"type": "string",
"format": "date-time",
"description": "Only what is new since this time (ISO 8601, for example \"2026-11-09T07:00:00Z\"). Pass the next_since value from your previous search_news reply. Overrides since_days. Stories are marked new or updated."
},
"detail": {
"enum": [
"brief",
"full"
],
"type": "string",
"description": "Less or more text. brief (default): every story with every article (publisher, date, link), the lead article's excerpt and the names in it. full: the same, plus every article's excerpt and a summary of each Polymath report."
},
"profile": {
"type": "string",
"maxLength": 600,
"description": "Who you are working for, in one or two sentences, for example \"a mid-size US law firm with a technology practice\". The profile tailors reports. It rarely selects stories by itself, so put the topics in \"interests\"."
},
"interests": {
"type": "array",
"items": {
"type": "string",
"maxLength": 120,
"minLength": 2
},
"maxItems": 8,
"description": "Three to eight specific topics this reader cares about, which you infer from the profile, for example [\"AI regulation\", \"data centers\", \"Nvidia\"]. Name each topic the way a news desk would. The name of a person or a company works as an interest too. Each interest gets its fair share of the results, and every story says which interests it matches and how well, so specific topics work far better than a profile alone."
},
"since_days": {
"type": "integer",
"maximum": 90,
"minimum": 1,
"description": "How many days back to look. Default 7 for stories and 30 for reports. Stories go back at most 30 days, reports 90."
},
"max_results": {
"type": "integer",
"maximum": 20,
"minimum": 1,
"description": "Most stories and most reports to return, each. Default 8."
}
},
"additionalProperties": false
}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 12, 2026.