Pressroom is a hosted platform with a web dashboard, a REST API, and an MCP server for headless access via Claude Code.

Architecture

System overview

Pressroom is a hosted platform with a web dashboard, a REST API, and an MCP server for headless access via Claude Code.

┌──────────────────────────────────────────────────────────────┐
│                       Web Dashboard                          │
│  Scout | Stories | Queue | Publish | Audit | Voice | Video   │
│  GSC | Blog | Assets | Scoreboard | Activity Log             │
└─────────────────────────┬────────────────────────────────────┘
                          │
┌─────────────────────────┴────────────────────────────────────┐
│                    Pressroom Platform                          │
│                                                               │
│  ┌──────────┐  ┌───────────┐  ┌───────────────────────────┐  │
│  │ API      │  │ Engine    │  │ Skills                    │  │
│  │          │→ │           │→ │ humanizer, seo_geo, taste │  │
│  │          │  │           │  │ channels/*, custom...     │  │
│  └──────────┘  └─────┬─────┘  └───────────────────────────┘  │
│                       │                                       │
│  ┌────────────────────┴──────────────────────────────────┐   │
│  │                  Service Layer                          │   │
│  │  services/performance.py — Content performance metrics │   │
│  │  services/blog_publisher.py — Blog publishing          │   │
│  │  services/storage.py — S3/cloud storage                │   │
│  │  services/social_auth.py — Social OAuth helpers        │   │
│  │  skills/channels/ — Channel generation templates       │   │
│  └────────────────────┬──────────────────────────────────┘   │
│                       │                                       │
│  ┌────────────────────┴──────────────────────────────────┐   │
│  │                   Data Layer                            │   │
│  │   Per-org isolated storage + DreamFactory               │   │
│  └────────────────────────────────────────────────────────┘  │
└──────────────────────────────────────────────────────────────┘
                          │
┌─────────────────────────┴────────────────────────────────────┐
│                    External Services                           │
│  Claude API | GitHub | LinkedIn | HubSpot | RSS | Medium      │
│  Google Search Console | Slack | Dev.to | YouTube | S3        │
└─────────────────────────┬────────────────────────────────────┘
                          │
┌─────────────────────────┴────────────────────────────────────┐
│               Remotion Renderer (React/Node)                  │
│  YouTube scripts → Remotion components → rendered video       │
└─────────────────────────┬────────────────────────────────────┘
                          │
┌─────────────────────────┴────────────────────────────────────┐
│                       MCP Server                              │
│  ~100 tools — headless access for Claude Code                 │
└──────────────────────────────────────────────────────────────┘

Data model

Core

Concept	Purpose
Organizations	Multi-tenant workspaces — each org is isolated
Signals	Industry signals pulled from HN, Reddit, GitHub, RSS
Content	Generated content queue with status workflow
Stories	Editorial story containers (curated signals + angle)
Story signals	Signal-story relationships with editor notes
Briefs	Daily brief summarizing signals + editorial angle
Settings	Key-value org configuration

SIGINT pipeline

Concept	Purpose
Sources	Global source library (Reddit, HN, RSS, X search, trends)
Org sources	Per-org source subscriptions
Raw signals	Crawled signals before relevance filtering
Org fingerprints	Per-org embedding vector for relevance scoring

Wire

Concept	Purpose
Wire sources	Company-owned feeds (GitHub repos, blog RSS, changelog)
Wire signals	Signals from company’s own channels

Supporting

Concept	Purpose
Team members	Discovered team members with expertise tags and voice analysis
Blog posts	Scraped blog context for voice training
Company assets	Tracked digital assets (logos, images, brand materials)
Site properties	SEO properties (domains, blog URLs) for auditing workflows
Audit results	SEO/README audit results
SEO PR runs	SEO PR pipeline run tracking
Email drafts	Composed email templates
AI visibility	Questions + citation results from ChatGPT/Perplexity/Claude
Competitor audits	Competitor SEO + citation scores
Token usage	API token consumption tracking by operation and over time
Activity log	War room log of all platform actions
Data sources	Custom external data source connections (MCP, REST API)
GSC data	Google Search Console analytics, URL inspection, sitemaps
Brand assets	Auto-discovered brand colors, logos, assets from domain crawl
Scoreboard	Multi-org rankings and per-team-member activity stats

Content status workflow

generating → queued → approved → published → tracked
                  ↘ spiked

generating — Claude is producing content
queued — Ready for human review
approved — Greenlit for publish
spiked — Rejected (used as anti-pattern memory for future generation)
published — Sent to destination platform
tracked — Post-publish performance metrics (views, likes, comments, shares) collected from LinkedIn, Dev.to

Multi-tenancy

Each organization is fully isolated. Signals, content, settings, team members, and audit results are scoped per org. Account-level settings (API keys, OAuth credentials) are shared across orgs within the same account.

Signal types

Type	Source
`github_release`	GitHub releases API
`github_commit`	GitHub commits API
`hackernews`	HN search API
`reddit`	Reddit scraping
`rss`	RSS feed parsing
`web_search`	Web search
`trend`	Trend discovery
`support`	Support ticket analysis
`performance`	Performance metrics
`gsc`	Google Search Console data

Content channels

Channel	Output format
`linkedin`	Professional post with hooks
`devto`	Dev.to article with frontmatter
`x_thread`	Thread-format short posts
`facebook`	Social post
`blog`	Long-form article with headers
`release_email`	Email template with subject line
`newsletter`	Newsletter digest format
`yt_script`	YouTube script with hook, sections, lower thirds
`github_gist`	GitHub gist (generated and published for team members)

AI models

Task	Model
Content generation	Claude Sonnet
Humanization	Claude Sonnet
Relevance filtering	Claude Sonnet
SEO analysis	Claude Sonnet
Fast tasks (classification)	Claude Haiku
Embeddings (optional)	Voyage

Access methods

Method	Best for
Web dashboard	Day-to-day content workflow, approvals, publishing
REST API	Custom integrations, automation, programmatic access
MCP server (~100 tools)	Claude Code sessions, AI-driven content operations