Docs

Connect PressRadar via MCP.

PressRadar gives your AI assistant tools via the Model Context Protocol (MCP) to search your sources and maintain your own journalist data.

Overview

The MCP server is available at the /api/mcp path of this website (Streamable HTTP). Every request authenticates with your workspace API key as a bearer token.

Your assistant decides when to use which tool. PressRadar never sends messages on its own; for regular briefings, set up a schedule in your client.

Good to know

MCP does not push news into an inactive conversation. Your AI assistant only queries PressRadar when you (or an automation) send a request. Automatic briefings need a separately configured schedule, such as a scheduled task in your AI client, if it supports one.

Setup

Sign up and complete the questionnaire.
Choose up to 10 sources and wait until at least one is active.
Create your API key in the workspace and copy it – it is shown only once.
Add PressRadar as a server in the settings or MCP configuration of Claude or Codex, and store the key there.
Ask your assistant, e.g. “Which relevant stories have appeared in my sources since yesterday?”

Client-specific instructions are available in your workspace after signing in.

Never paste your API key into an AI chat. Set it as an environment variable (PRESSRADAR_API_KEY) or in your client's dedicated settings field. If a key was shared by accident, revoke it and create a new one. Never paste your API key into a chat. It belongs in your client’s settings or configuration file. We only store a hash, and you can revoke the key at any time.

Example questions

Read-only“What stories relevant to my company appeared in the latest collection?”
Read-only“Which journalists covered this topic?”
Read-only“Show the sources and explain why this story is relevant.”
Writes to your workspace“Save this journalist to my list and note that she covers energy policy.”
Writes to your workspace“Log that I emailed this journalist today and sent our press release.”

Instructions per client

Researched against each vendor's official documentation. The status shows honestly what works directly and what needs a workaround.

Claude CodeSupportedNot tested end-to-end

Claude Code supports remote HTTP MCP servers with custom headers. The API key is sent as an Authorization header.

Set your PressRadar API key as the environment variable PRESSRADAR_API_KEY in your shell (e.g. in ~/.zshrc or ~/.bashrc).
Option A: run `claude mcp add`. Your shell inserts the key and Claude Code stores it in its local configuration.
Option B (for shared projects): create a `.mcp.json` file in the project folder. It only contains `${PRESSRADAR_API_KEY}`; Claude Code expands it from your environment at startup, so the key never lands in the repository.
Restart Claude Code and check the connection with `/mcp`.

Add via CLI

claude mcp add --transport http pressradar https://pressradar.website.cloud.brutal.ai/api/mcp \
  --header "Authorization: Bearer $PRESSRADAR_API_KEY"

.mcp.json in your project

{
  "mcpServers": {
    "pressradar": {
      "type": "http",
      "url": "https://pressradar.website.cloud.brutal.ai/api/mcp",
      "headers": {
        "Authorization": "Bearer ${PRESSRADAR_API_KEY}"
      }
    }
  }
}

Based on official docs as of 2026-11-06; not yet tested end-to-end.

Sources

https://code.claude.com/docs/en/mcp

Claude DesktopWorkaroundNot tested end-to-end

Claude Desktop's connector settings are the same as claude.ai's: a fixed API key only works there with the "Request headers" beta. As a workaround, you can add PressRadar as a local server through the open-source tool mcp-remote.

If your organization has the "Request headers" beta, follow the Claude on the web instructions. The setting then applies in Claude Desktop too.
Otherwise, install Node.js so that `npx` is available.
Open Claude Desktop's `claude_desktop_config.json` file and add the entry below.
Replace `$PRESSRADAR_API_KEY` with your key directly in that file. Claude Desktop does not read environment variables from your shell here.
Keep `args` free of spaces around the colon (`Authorization:${AUTH_HEADER}`). According to mcp-remote, spaces in `args` are passed incorrectly on Claude Desktop for Windows.
Fully restart Claude Desktop.

claude_desktop_config.json (via mcp-remote)

{
  "mcpServers": {
    "pressradar": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://pressradar.website.cloud.brutal.ai/api/mcp",
        "--header",
        "Authorization:${AUTH_HEADER}"
      ],
      "env": {
        "AUTH_HEADER": "Bearer $PRESSRADAR_API_KEY"
      }
    }
  }
}

Based on official docs as of 2026-11-06; not yet tested end-to-end.

Sources

Claude (claude.ai)Not directly yetNot tested end-to-end

Custom connectors in claude.ai support OAuth or servers without sign-in. Fixed API keys via "Request headers" are a beta for a limited set of organizations. PressRadar does not offer OAuth yet, so the connection only works if your organization has this beta.

In claude.ai, open "Customize > Connectors" and click "Add custom connector". On Team and Enterprise plans, an Owner adds it under "Organization settings > Connectors" (Add > Custom > Web).
Enter the address shown below as the MCP server URL.
Under "Authentication", choose "No sign-in".
Open "Request headers", pick the header name `authorization` and enter `Bearer ` followed by your key (with the space) as the value. Alternatively: header `x-api-key` with the bare key as the value.
If there is no "Request headers" section, your organization does not have the beta yet. Use Claude Code or Claude Desktop with mcp-remote instead.
Note: on Team and Enterprise plans every member shares the stored key. Headers cannot be edited after saving; remove the connector and add it again to change them.

MCP server URL

https://pressradar.website.cloud.brutal.ai/api/mcp

Request header (authorization)

authorization: Bearer $PRESSRADAR_API_KEY

Based on official docs as of 2026-11-06; not yet tested end-to-end.

Sources

OpenAI Codex CLISupportedNot tested end-to-end

Codex supports Streamable HTTP MCP servers and reads the bearer token from an environment variable. The key itself is never written to the config file.

Set your PressRadar API key as the environment variable PRESSRADAR_API_KEY.
Run `codex mcp add`. You pass only the variable's name, not its value.
Alternatively, add the server directly to `~/.codex/config.toml` (or `.codex/config.toml` in a trusted project).
Restart Codex. `codex mcp --help` shows the commands for listing and removing servers.

Add via CLI

codex mcp add pressradar --url https://pressradar.website.cloud.brutal.ai/api/mcp \
  --bearer-token-env-var PRESSRADAR_API_KEY

~/.codex/config.toml

# ~/.codex/config.toml
[mcp_servers.pressradar]
url = "https://pressradar.website.cloud.brutal.ai/api/mcp"
bearer_token_env_var = "PRESSRADAR_API_KEY"

Based on official docs as of 2026-11-06; not yet tested end-to-end.

Sources

OpenAI Codex (IDE extension)SupportedNot tested end-to-end

The Codex IDE extension shares its configuration with the Codex CLI (`~/.codex/config.toml`). A server you set up once is available there too.

Set up PressRadar as for the Codex CLI, with `codex mcp add` or directly in `~/.codex/config.toml`.
Make sure PRESSRADAR_API_KEY is set in the environment your editor is launched from.
Reload the editor window or restart the extension.

~/.codex/config.toml

# ~/.codex/config.toml
[mcp_servers.pressradar]
url = "https://pressradar.website.cloud.brutal.ai/api/mcp"
bearer_token_env_var = "PRESSRADAR_API_KEY"

Based on official docs as of 2026-11-06; not yet tested end-to-end.

Sources

Tool reference

Parameter descriptions come straight from the server's validation schema.

News & sources

`list_sources`Read-only

List the publisher websites this workspace monitors, with collection health (last success, next run, failures, articles in the last 7 days, stale flag) and the plan's source limit.

No parameters.

Returns
sources[]: sourceId, publisherId, name, domain, status, lastSuccessAt, nextCollectAt, consecutiveFailures, lastRunStatus, articles7d, stale
limit / used: Source slots on the plan and how many are in use

Example

list_sources({})
→ {"sources":[{"name":"Handelsblatt","status":"active","lastSuccessAt":"2026-11-06T06:02:11Z","stale":false}],"limit":10,"used":4}

`search_news`Read-only

Search stories (clusters of articles about the same event) from this workspace's sources, ranked by the workspace's relevance score, then recency. Returns the top article, outlets, authors, matched keywords, short relevance reasons and a labelled model summary. Paginate with nextCursor.

Parameters: search_news
Parameters	Type	Description
query	string	Words to find in title, teaser or text (all must match; max 8 terms; "quoted phrase" allowed). Omit to list the most relevant stories.
since	string	Only articles published/discovered at or after this time (ISO 8601, e.g. 2026-11-01 or 2026-11-01T08:00:00Z)
until	string	Only articles published/discovered at or before this time (ISO 8601, e.g. 2026-11-01 or 2026-11-01T08:00:00Z)
sinceHours	integer	Shortcut for since: only the last N hours (e.g. 6 for the latest collection, 24 for a day)
publisherIds	string[]	Limit to these sources (publisherId from list_sources)
minScore	number	Minimum relevance score 0..100 for this workspace
includeExcluded	booleandefault: false	Also return articles that matched an exclusion rule
author	string	Only articles with an author whose name contains this text
cursor	string	nextCursor from the previous result, to get the next page
limit	integerdefault: 10	Stories per page (1..25)

Returns
stories[]: storyId, title, score (0..100 or null), topArticle {url, title, outlet, publishedAt, teaser}, articleCount, outlets, authors, matchedKeywords, relevanceReasons, modelSummary
nextCursor: Pass as cursor for the next page; null when done
freshness: When the latest collection ran and when the next is scheduled

Example

search_news({"query":"Wasserstoff Förderung","sinceHours":24,"limit":5})
→ {"stories":[{"storyId":"cm…","title":"Bund stockt Wasserstoff-Förderung auf","score":87,"articleCount":4}],"nextCursor":null}

`get_story`Read-only

Get one story with its articles (URL, outlet, dates, teaser, optional full text), authors merged with your own journalist notes and publicly evidenced contacts (with source URL), the workspace's relevance assessment, and an explicit list of missing data.

Parameters: get_story
Parameters	Type	Description
storyIdrequired	string	storyId from search_news
includeText	booleandefault: false	Include the article full text (third-party content, up to 20,000 characters per article)
maxArticles	integerdefault: 10	How many articles of the story to return (1..20)

Returns
articles[]: url, title, outlet, publishedAt, teaser, text/truncated (with includeText), textSource, textStatus, authors[] with merged journalist, missing[]
relevance: score, reasons, matchedKeywords, method, modelSummary (labelled)
missing[]: Data not available, e.g. publishedAt, authors, text, contacts

Example

get_story({"storyId":"cm…","includeText":true,"maxArticles":3})
→ {"title":"…","articles":[{"outlet":"FAZ","textStatus":"full","textSource":"publisher","truncated":false}],"missing":["contacts"]}

`get_freshness`Read-only

When the latest scheduled collection finished, when the next one is due (every 6 hours), and which sources are stale, partial or failing.

No parameters.

Returns
lastCollectionAt / nextScheduledAt: ISO timestamps
staleSources / partialSources / failedSources: Sources with problems and the reason

Example

get_freshness({})
→ {"lastCollectionAt":"2026-11-06T06:04:00Z","nextScheduledAt":"2026-11-06T12:00:00Z","staleSources":[]}

Journalists & your own data

`search_journalists`Read-only

Find journalists by name, outlet or topic. Ranked by how many matching articles they wrote in your sources in the last 90 days. Your own records and notes take priority over public data.

Parameters: search_journalists
Parameters	Type	Description
query	string	Name (or part of it)
topic	string	Topic words; ranks journalists by recent coverage (90 days) in your sources
outlet	string	Outlet name or domain
limit	integerdefault: 10	Max results (1..25)

Returns
journalists[]: workspaceJournalistId, journalistId, name, outlet, role, notes, tags, doNotContact, contacts[] (origin workspace / public_source), matchingArticles90d

Example

search_journalists({"topic":"Batteriezellen","limit":5})
→ {"journalists":[{"name":"…","outlet":"Handelsblatt","matchingArticles90d":6}]}

`get_journalist`Read-only

One journalist: your own record (notes, tags, contacts, interactions) first, then publicly evidenced contacts with source URL, and their recent articles in your sources.

Parameters: get_journalist
Parameters	Type	Description
workspaceJournalistId	string	Your workspace's own record id
journalistId	string	Global journalist id (from get_story / search_journalists)

Returns
contacts[]: origin 'workspace' (yours, with contactId) first, then 'public_source' with sourceUrl/sourceType
recentArticles[] / interactions[]: Latest coverage and your logged interactions

Example

get_journalist({"journalistId":"cm…"})
→ {"name":"…","contacts":[{"origin":"public_source","kind":"email","sourceUrl":"https://…/autor/…"}]}

`save_journalist`Writes to your workspace

Create or update your workspace's own record for a journalist (private to your workspace). Link it to a known journalist with journalistId. Only the fields you pass are changed; notes and tags are replaced.

Parameters: save_journalist
Parameters	Type	Description
workspaceJournalistId	string	Update this existing workspace record
journalistId	string	Global journalist id to link (from get_story / search_journalists)
name	string	Full name (required when creating without journalistId)
outlet	string	Outlet / publication
role	string	Role or beat, e.g. 'Wirtschaftsredakteurin, Energie'
notes	string	Your private notes (replaces previous notes)
tags	string[]	Tags (replaces previous tags)
doNotContact	boolean	Mark as do-not-contact

Returns
created / journalist: Whether a record was created, and the merged record incl. workspaceJournalistId

Example

save_journalist({"journalistId":"cm…","role":"Energie-Ressort","notes":"Bevorzugt Exklusivdaten, nicht vor 10 Uhr anrufen","tags":["energie"]})
→ {"created":true,"journalist":{"workspaceJournalistId":"cm…"}}

`add_journalist_contact`Writes to your workspace

Add a contact detail to your workspace's journalist record (email, phone, social handle, …). Validated; stored privately.

Parameters: add_journalist_contact
Parameters	Type	Description
workspaceJournalistIdrequired	string	Workspace record id from save_journalist
kindrequired	email \| phone \| website \| profile \| x \| linkedin \| bluesky \| mastodon \| instagram \| threads \| other	Contact type
valuerequired	string	Email, phone, URL or handle
label	string	e.g. 'work', 'newsroom'
source	string	Where this contact came from (e.g. 'email signature', a URL)

Returns
contactId: Id for remove_journalist_contact

Example

add_journalist_contact({"workspaceJournalistId":"cm…","kind":"email","value":"name@example.org","source":"E-Mail-Signatur"})
→ {"contactId":"cm…","kind":"email"}

`remove_journalist_contact`Writes to your workspace

Remove one of your workspace's own contact entries. Public contacts cannot be removed here.

Parameters: remove_journalist_contact
Parameters	Type	Description
contactIdrequired	string	contactId of a workspace contact (origin 'workspace' in get_journalist)

Returns
removed: true when deleted

Example

remove_journalist_contact({"contactId":"cm…"})
→ {"removed":true}

`log_interaction`Writes to your workspace

Record a pitch, email, call, meeting, coverage or note with a journalist in your workspace.

Parameters: log_interaction
Parameters	Type	Description
workspaceJournalistIdrequired	string	Workspace record id from save_journalist
kindrequired	email_sent \| email_received \| call \| meeting \| pitch \| coverage \| note	Interaction type
occurredAt	string	When it happened; default now (ISO 8601, e.g. 2026-11-01 or 2026-11-01T08:00:00Z)
subject	string	Short subject
summary	string	What happened / was agreed
articleId	string	Related article (e.g. resulting coverage)

Returns
interactionId: Id of the logged entry

Example

log_interaction({"workspaceJournalistId":"cm…","kind":"pitch","subject":"Studie Q4","summary":"Interesse, Rückmeldung bis Freitag"})
→ {"interactionId":"cm…","kind":"pitch"}

`request_journalist_enrichment`Writes to your workspace

Queue a background lookup of publicly listed contact details for a journalist (author pages). Returns 'queued' immediately; results appear in get_journalist after a later run. Limited to 20 per workspace per day.

Parameters: request_journalist_enrichment
Parameters	Type	Description
journalistIdrequired	string	Global journalist id (from get_story / search_journalists)

Returns
status: queued | already_queued, remainingToday

Example

request_journalist_enrichment({"journalistId":"cm…"})
→ {"status":"queued","remainingToday":19}

Start free FAQ

Overview

Setup

Example questions

Instructions per client

Tool reference

News & sources

list_sourcesRead-only

search_newsRead-only

get_storyRead-only

get_freshnessRead-only

Journalists & your own data

search_journalistsRead-only

get_journalistRead-only

save_journalistWrites to your workspace

add_journalist_contactWrites to your workspace

remove_journalist_contactWrites to your workspace

log_interactionWrites to your workspace

request_journalist_enrichmentWrites to your workspace

`list_sources`Read-only

`search_news`Read-only

`get_story`Read-only

`get_freshness`Read-only

`search_journalists`Read-only

`get_journalist`Read-only

`save_journalist`Writes to your workspace

`add_journalist_contact`Writes to your workspace

`remove_journalist_contact`Writes to your workspace

`log_interaction`Writes to your workspace

`request_journalist_enrichment`Writes to your workspace