innie

Project Description

Innie is an agentic assistant that runs your LinkedIn content workflow end to end. It helps you draft posts, test angles, and finalize a version that is ready to post without sounding like AI slop. It learns your style over time and optimizes based on what performs.

The problem we’re solving and its impact:

Screen fatigue is real. In 2025, 50% of adults said they actively limited their social media usage. But consistency still pays. LinkedIn rewards “show up every week,” and 34% of recruiters say thought leadership can raise salary offers by up to 10%.

Plus, we love the show Severance. We wanted an “innie” to handle the LinkedIn persona so the “outie” can stay focused on actual life. Coffee shops, dog walks, hackathons :)

Key technical architecture and components

1. Web app (UI): Users talk to Innie, review drafts, and approve final posts.
2. Node.js API: Orchestrates the agent workflow and routes tasks to the right sub-agent.
3. Voice agent (storytelling): turns a short conversation into a structured post in your voice.
4. Research agent: pulls relevant context and stats, then blends them into the draft when useful.
5. Lookalike audience agent: compares the draft to patterns from high-performing posts in similar niches and rewrites.
6. Consistency engine: schedules reminders (email or Slack) so you keep showing up weekly.
7. Memory + storage: Stores your preferences, drafts, and performance signals so innie improves over time.

Sponsor tools / protocols used and how we integrated them

1. ElevenLabs (voice): Powers the voice experience so users can talk to Innie instead of typing, making story capture fast and natural.
2. CodeRabbit (code review): Used for automated PR feedback to speed up iteration and keep quality high while building fast.

Built With

• coderabbit
• elevenlabs
• javascript
• node.js
• openai