Echo Harbor Case Study
Built a Claude-Powered Platform That Surfaces the Claims, Sources, and Bias in Content.
About the client
Echo Harbor is an AI software company focused on digital media and news analysis. Its platform helps users capture content from across articles, videos, podcasts, and other sources and turn it into structured, source-traceable insights.
The challenge
Readers face more news than they can process, across articles, PDFs, videos, podcasts, and emails, with no fast way to see what a story says, who it affects, or where it is biased. The client wanted one platform to capture any format and return grounded analysis, claims, sources, bias, and the stakeholders behind each story.
Our solution
3 products that run on 1 platform, sharing a single capture-and-analysis engine built on Anthropic’s Claude models through the EML layer:
WikiClip
(a knowledge workspace; web released Jul 17, 2026; iOS app submitted to the App Store Jul 20, 2026).Echo Harbor News
(a bias-scored news site).AIThinks
(an LLM that shows its reasoning).What the product does
Captures any format (articles, PDFs, images, emails, audio, video), transcribing and indexing audio and video so everything is searchable.
Analyzes any clip through grounded Lenses: key facts and claims, notable quotes, key sections, citations, summaries at 3 lengths, and bias or emotional tone, each quoted from the source.
Automates repeat analysis: users define the checks they want and apply them to a whole folder, so every new clip is analyzed the moment it lands.
Maps the stakeholders in each story: who benefits, who suffers, and who was left out.
Results
Echo Harbor News holds 181K+ articles across 20 topics, now expanding into local news and growing by a few thousand a day.
Early SEO traction: ranking for 66 keywords in July (+13 month over month)
Around 30 two-week sprints delivered; final sprint closed Jul 22, 2026.
Team scaled to 10–19 engineers during the build.
- FastAPI-based microservice infrastructure designed for enterprise-scale AI workloads.
- Layered safety, meaning, and modulation services that keep AI responses controlled and reliable.
- Claude Code is the development backbone: every engineer works in a mandatory Claude account, using it to accelerate implementation, unit testing, and deployment.
- The team pivoted from Traditional Dev Tools to AI Native Tools.