DevSpace

Landing Page
Call Mode
Github Connection

Introduction DevSpace is the future of developer productivity: an intelligent engineering collaboration platform with AI agents that speak, listen, analyze your codebase, and take real, actionable steps—like reviewing, refactoring, and creating GitHub pull requests—all in real time. Inspiration Building and scaling software is hard. Developers spend up to 40% of their time waiting for reviews, switching contexts, and debugging. As a founder and engineer, I wanted to work with a “real” engineering team—instantly available and superhumanly efficient. DevSpace was born out of this need: AI engineer agents that join your calls, understand your code, and actually ship improvements. What It Does • Voice/Video Collaboration: Talk to multiple AI “engineers” who respond with natural voices (FishAudio), video avatars (Tavus), and specialty knowledge. • Direct GitHub Integration: Connect your repo, and let agents review, analyze, and propose changes on the actual codebase. • Code Review & Editing: Agents provide line-by-line feedback, fix bugs, refactor code, and create automated pull requests. • Task Management: Assign tasks, track progress, and receive notifications from engineer agents. • Custom Specialties: Backend, frontend, security, DevOps engineers—each agent has a unique persona and domain expertise. • Works in Real Time: No waiting—responses, reviews, and code suggestions happen as you speak. Tech Stack • Next.js + React (UI engine) • FishAudio (STT/TTS, custom voice agents) • Tavus (Video avatars for engineers) • Supabase (database, user auth, task state) • GitHub API (repo access, PR automation) • Windmill (workflow orchestration—modular, for advanced work) • Google Gemini & OpenRouter (LLM internals demoed) • Custom multi-agent orchestration (voice, text, persona management) Challenges We Ran Into • Real-time voice streaming with multiple agent personas • Integrating true GitHub write access safely • Keeping UI/UX beautiful while adding next-gen features • Synthesizing voices for “AI engineers” from reference samples • Context switching between hackathon tools! Accomplishments We’re Proud Of • Built, refactored, and shipped full-stack, voice-driven AI engineers in under 48 hours • Custom voice and video AI agent demos • GitHub PR creation from AI agents—truly autonomous code improvement • Multi-agent simultaneous “FaceTime” for devs What’s Next For DevSpace • Production-grade deployment with more agent types (AI/ML, QA, Data) • Advanced context awareness: agents learn your project • Real-time feedback (not just suggestions—actual merged PRs!) • Open source for the developer community • Commercial release for teams, startups, and solo founders