Everything your agent needs to do PR

the media intelligence layer for your agent

Give your agent live news, the journalists actually covering it, and a drafted angle for each — and keep every pitch behind human sign-off.

Get an API keyRead the docs

Free, no credit card required.

Install in 30 seconds
# add Medialyst as a remote MCP server
export MEDIALYST_API_KEY="<YOUR_API_KEY>"
AUTH_HEADER="Authorization: Bearer $MEDIALYST_API_KEY"

claude mcp add \
  -t http -s user \
  -H "$AUTH_HEADER" \
  medialyst https://medialyst.ai/api/mcp

claude mcp list
Choose your harness. Copy the setup. Keep the key out of prompts.
Supports any agent harness via MCP
The primitives

Three primitives. One rule: you stay in charge.

Agents do their best work grounded in current reporting — and their worst when they improvise who to contact. Every Medialyst primitive keeps human judgment in the loop.

01 · Real-time news search

Search the news in real time.

Generic web search was never built for news — the cycle moves too fast and stale results miss the window. Medialyst runs a dedicated real-time news search and attaches the publication's metadata to every result — including whether it's original editorial or a newswire, plus authority and reach. That's context a plain search API never gives you.

  • Built for the news cycle live coverage over keywords, URLs, or topics — not a stale web index.
  • Editorial vs. newswire, classified know whether a result is original reporting or a press-release wire — plus authority and reach.
  • 0.1 credits per search shared across UI, REST, and MCP.
mcpsearch_newslive · 0.1 cr
// tbs "qdr:h" → only the last HOUR of news
search_news({
  q: "AI safety regulation europe",
  tbs: "qdr:h"
})
response
{
  "status": "success",
  "data": { "articles": [
    {
      "title": "EU lawmakers press AI safety timeline",
      "source": "Reuters",
      "metadata": {
        "publication_type": "editorial",
        "domain_authority": 94,
        "estimated_monthly_organic_traffic": 41000000
      }
    },
    {
      "title": "Vendor launches AI compliance toolkit",
      "source": "PR Newswire",
      "metadata": {
        "publication_type": "newswire",
        "domain_authority": 76,
        "estimated_monthly_organic_traffic": 8200000
      }
    }
  ] }
}
02 · Find the right journalists

Find journalists using natural language.

Describe the story in plain language. Medialyst spins up thousands of subagents on demand to resolve the journalists actually covering that beat right now — and hands back a media list in minutes, not days. Find the right people fast, at scale, so your agent moves at the speed of the news.

  • Plain-language discovery describe the beat; get the bylines actually covering it.
  • Thousands of subagents, in parallel research at scale instead of one byline at a time.
  • Shareable read-only views for human review.
mcpcreate_media_list32 journalists
// describe who you want — plain language
create_media_list({
  name: "AI policy reporters",
  source: {
    type: "prompt",
    prompt: "journalists covering AI policy at top US outlets"
  }
})
response
{
  "media_list": {
    "id": "list_8f2a",
    "name": "AI policy reporters",
    "journalists": 32,
    "rows": [
      { "name": "Maya Chen",
        "outlet": "The Information", "fit": 0.94 },
      { "name": "Rita Liao",
        "outlet": "TechCrunch", "fit": 0.89 }
    ]
  }
}
03 · Human-in-the-loop outreach

Keep final judgment with the person whose reputation is on the line.

A PR agent without grounded media intelligence can pitch the wrong beat, invent fit, and burn trust with journalists. Medialyst lets the agent draft and organize work, but keeps review, approval, and sending in the app.

  • No public send endpoint and no fake MCP send tool.
  • Preview before applying so prompts and outputs can be checked.
  • App-only Gmail/Outlook send keeps review explicit.
mcpapply_table_actiondraft only · no send
// "Draft a Gmail pitch for each journalist"
apply_table_action({
  action: "create_column",
  column: { id: "gmail_draft" }
})
response
{
  "column": "gmail_draft",
  "status": "draft_created",
  "draft": {
    "to": "[email protected]",
    "subject": "A data point for your...",
    "send": "app_only"
  }
}
Grounded, not guessed

Ground your agents in live, verified information — not pre-training data.

A vanilla agent answers from frozen training data and improvises the rest. The same agent working from a Medialyst media list reasons over live coverage and verified journalist records — every row backed by a real byline.

Without Medialyst
Vanilla agent
With Medialyst
Medialyst MCP
News search
Falls back to web search, which returns stale data
Real-time news search — specify the window an article was published in
Finding journalists
Reads a stale byline from memory
Dedicated byline and author-retrieval flow with waterfall email lookup, cross-referenced against our media database
Email validation
Does not validate — sends to whatever it guessed
Every email validated and verified in real time
Journalist beat understanding
Guessed from memory, often out of date
Resolved from real, recent bylines
Publication context
Unknown or invented
Outlet type, authority, and reach attached
Email sending
No guardrails — can burn your reputation spamming journalists
Human-in-the-loop, built in by design
Same model, same prompt. The only difference is whether the agent is working from memory or from live, verified media data.
The surface

Same key. REST or MCP. Thirteen tools through MCP.

The MCP server is self-documenting via get_usage_guide and get_tool_reference. REST exposes the same production media-list workflow contract at the documented API paths.

MCP Server

Drop into Claude Cowork, Cursor, Codex, or any agent harness.

REMOTEhttps://medialyst.ai/api/mcp

13 tools, self-documenting. Call get_usage_guide first and the model gets current operating instructions from the server instead of stale copied schemas.

get_usage_guideget_tool_referenceget_credit_balancesearch_newscreate_media_listlist_media_listsget_media_listinspect_tableread_full_valuespreview_column_renderapply_table_actioncreate_share_linkdelete_media_list
Streaming HTTP transport · structured tool errors · org-scoped keys
REST API

The same surface, hand-callable.

BASEhttps://medialyst.ai

API-key authenticated and scoped per organization. The paths below are the public v1 routes; outreach send and reply-fetch endpoints are intentionally not listed because they do not exist in public v1.

POST/api/v1/news/search
GET/api/v1/media-lists
POST/api/v1/media-lists
GET/api/v1/media-lists/:id
DELETE/api/v1/media-lists/:id
POST/api/v1/media-lists/:id/inspect
POST/api/v1/media-lists/:id/full-values
POST/api/v1/media-lists/:id/column-render-preview
POST/api/v1/media-lists/:id/actions
POST/api/v1/media-lists/:id/shares
Per-call credit cost · 402 on quota exhaustion · shared UI/API/MCP credits
Pricing

Credits, not seats. Agents don't have seats.

API keys and MCP are available on every plan. Credits are shared across UI, API, and MCP.

Starter
For solo marketers and founders running their own PR
$149/ month
3,500 credits / mo · 35K news calls
  • API + MCP access
  • 3 active media lists
  • 1,000 articles per campaign
  • Gmail/Outlook send in app
  • Shared credits across UI, API, MCP
Start with Starter
Most popular
Pro
For marketing teams running campaigns at scale
$299/ month
10,000 credits / mo · 100K news calls
  • Everything in Starter
  • Unlimited active media lists
  • 1,000 articles per campaign
  • No Medialyst branding on shared lists
  • Team workflows + priority support
Choose Pro
Scale
For agencies and platforms embedding Medialyst
$800/ month
50,000 credits / mo · 500K news calls
  • Everything in Pro
  • White-labeled shared lists
  • Custom onboarding and integrations
  • Dedicated support
  • Volume and procurement support
Talk to us
How credits are used

Every action has a flat, predictable cost.

Credits are shared across UI, API, and MCP and draw from the same workspace pool — so agent usage and in-app work bill the same way.

1 credit ≈ $0.043 at Starter · less at Pro and Scale
News search
Per real-time news query — specify the window an article was published in.
0.1 credit
Enrich a journalist
Complete contact information, including an email verified in real time.
1 credit
Find recent articles
Pull a journalist's most recent bylines to ground beat and angle work.
1 credit
AI analysis
Beat matching and pitch-angle suggestions for each journalist.
1 credit
Custom research agent
Deep custom research — final cost varies by the model you run.
From 5 credits
Need bespoke volume, custom retention, or an MSA? Talk to us →
FAQ

Questions procurement asks. Answers agents can read.

Our search results mirror search-engine news search, so you get the same real-time coverage you would expect. When a source is in our publication index, the news item also gets Medialyst metadata such as publication type, domain authority, and estimated monthly organic traffic.

Journalist rows and contact enrichment happen when you create a media list from selected articles, URLs, or seed keywords. Search is retrieval plus source classification, not a fake byline-resolution endpoint.

Yes. Add it as a remote HTTP MCP server with your API key. The install snippet at the top of this page matches the documented claude mcp add --transport http command.

Journalist data is retrieved in real time, so everything stays fresh. Contacts and bylines are resolved against live sources at request time rather than served from a stale snapshot.

We log requests for billing, abuse prevention, and debugging. We do not train models on your media lists, queries, or outreach content.

The REST API returns 402 Payment Required. MCP returns the same condition as structured tool content so the agent can surface the quota issue instead of silently continuing.

Yes, inside the same organization and product workflow. Reselling raw journalist contact data is not permitted; building an agent or product experience on top of Medialyst data is what the API is for.

Muck Rack, Cision, and Prowly are built around human PR teams using a UI — they sell a static database with a per-seat license.

Medialyst is a callable surface. The journalist data is resolved in real time from any article, not pulled from a fixed database. Every primitive — search, list creation, table action, share link — is callable from REST or MCP. Pricing is per credit, not per seat, so agent usage draws from the same shared workspace pool as the app.

If you're a PR pro at a desk, those tools are mature and well-supported. If you're a marketer driving work through an agent, Medialyst is the layer underneath that work.