Cadence — AI Content Autopilot for LinkedIn & Blog
GOLLE IT in-house SaaS: agentic AI content autopilot with a 7-phase pipeline, multi-workspace architecture and direct publishing to LinkedIn, blog and WordPress.

7
Pipeline phases
3+
Channels (LinkedIn, blog, WordPress)
Multi-Tenant
Workspace isolation
DSGVO
EU hosting & compliant
About the Client
Cadence is an in-house SaaS product by GOLLE IT — an agentic AI content autopilot that automates the entire content pipeline from topic discovery through research and generation to scheduled multi-channel publishing. Target audience: solo creators, B2B companies and especially agencies that need to serve multiple clients in parallel, each with their own brand voice. Available as Free, Starter, Pro and Teams plans at cadence.golle.io.
The Challenge
Even with ChatGPT, Claude or Gemini, content owners today still spend several hours per content piece. The problem is not the writing itself — it is everything around it: finding topics, agentic research, platform-specific adaptation (LinkedIn character limits, blog SEO, WordPress multi-site), translation, and publishing at the right moment. For agencies serving multiple clients, the problem multiplies, because each client has its own brand voice, its own industry topics and its own channels.
A tool that is multi-tenant by design AND truly agentic — not just prompt-hopping in ChatGPT — was missing from the market.
Our Solution
We built Cadence as an end-to-end SaaS with a 7-phase pipeline: Ideation → Analysis → Research → Creating → Review → Translation → Publishing. The research agent is built on LangGraph as a multi-step reflection loop: it plans queries, searches the web via Tavily, curates sources, detects gaps and re-researches — like a senior researcher. The LLM layer is fully swappable via OpenRouter: Claude Sonnet/Opus 4.6, Gemini 2.5 Pro, GPT-4o or self-hosted Llama models run without code changes; workspaces can even bring their own OpenRouter keys.
Each workspace has its own writing-style profile (extracted from 200–15,000 characters of sample text), its own industry feeds and its own channel configuration — ideal for agencies serving multiple clients. Direct publishing reaches LinkedIn (personal and company pages, OAuth 2.0), WordPress (multi-site via Application Passwords + REST API v2) and the blog. Images are generated via Google Gemini 3 Pro (default, configurable) in platform-appropriate aspect ratios. SSE streams deliver live updates for every pipeline step; background jobs keep running even if the browser disconnects. Stripe Subscriptions + credits system, S3-compatible object storage (MinIO), Sentry monitoring, Playwright for website analysis.
Solution Highlights
- Agentic 7-phase pipeline with live status and human-in-review loop
- Multi-tool research agents with reflection loops (not a simple Google-search wrapper)
- LLM-agnostic: automatic upgrades to the strongest current frontier model
- Writing-style profiles per workspace — posts still sound like YOU, not like ChatGPT
- Direct publishing without platform workarounds: LinkedIn, WordPress (multi-site), blog
- Multi-workspace architecture for agencies serving many clients
The Results
Cadence runs as a productive SaaS with Free, Starter (€29/month), Pro (€79/month) and Teams plans at cadence.golle.io. GOLLE IT itself uses Cadence in-house for the blog pipeline on golle-it.de — so the platform is not a lab experiment but dogfooded SaaS. The LLM layer is continuously upgraded to the strongest available frontier models; the channel list expands step by step.
Cadence demonstrates the depth of GOLLE IT's in-house expertise in agentic AI, multi-tenant SaaS architecture and LLM-agnostic pipelines.
What We Built
Discovery agent
Suggests fresh topics daily — based on profile, industry and current trends. Refinable in interactive chat, with a quality score (1–10) per idea.
Multi-tool research agent
Plans queries, scans the web with multiple tools, evaluates sources, detects gaps and re-researches. Delivers a structured briefing with all sources — not a simple ChatGPT output.
Per-workspace style training
200–15,000 characters of sample text are enough — Cadence extracts tone, syntax, vocabulary, signature phrases and do/don't rules. For agencies: a separate profile per client.
Multi-channel generation in one run
Blog article, LinkedIn post and AI-generated hero image (in the correct aspect ratio) emerge from a single pipeline run. "Fix with AI" shortener trims overlong LinkedIn posts without losing hook or CTA.
Per-channel smart scheduling
Direct publishing to LinkedIn (personal and company pages), WordPress (multi-site) and blog. Smart scheduling proposes the optimal weekday per channel.
Multi-workspace + REST API
Per workspace: own discovery, own style profiles, own channels. Full REST API with OpenAPI docs and SSE streams for live updates from the Pro plan onwards. EU hosting, GDPR-compliant.
Project Insights
Visit live projectIllustrative, AI-generated depiction of the interface – not the actual application. You can see the real interface at cadence.golle.io.


