Devpost
Participate in our public hackathons
Devpost for Teams
Access your company's private hackathons
Grow your developer ecosystem and promote your platform
Drive innovation, collaboration, and retention within your organization
By use case
Blog
Insights into hackathon planning and participation
Customer stories
Inspiration from peers and other industry leaders
Planning guides
Best practices for planning online and in-person hackathons
Webinars & events
Upcoming events and on-demand recordings
Help desk
Common questions and support documentation
Sort by:
Never Miss the Opportunities You Deserve NextLane AI finds and filters jobs, scholarships, and hackathons tailored to your profile, so you don't have to search.
Turn electrical schematic images into structured, editable digital models with AI and human review.
Verified agent actions, safer by middleware.
We often "approve" AI agents without seeing what they’ll really do. AirLock changes that by isolating every run, blocking dangerous changes, and promoting only what it truly reviewed.
An AI tutor that answers only from your course notes, cites the exact page, and checks its math with SymPy. Powered by NVIDIA Nemotron 3 Nano, Super and Ultra on Nebius Token Factory.
An evidence-grounded Gemini agent for scientific reviewers: navigate complex climate documents, analyze passages, trace context, and draft human-reviewed comments without automatic submission.
WARDEN is a security firewall for AI-agent capabilities, using Gemini to detect malicious tools, Ed25519 to seal trusted ones, and runtime policy enforcement to keep agent fleets safe.
Give AI agents real power without giving them the final word. Release Sentinel lets agents act through WebMCP while a deterministic Gatekeeper retains release authority.
An advanced, offline-first Autonomous AI Code Writer and Multi-Agent Operating System powered entirely by local models
Professional Agents. Multi-agent analyser: upload paper+marks, map chapters, track a student all year. Teacher/student/parent views; ping teacher only if missing data or a judgment call.
Other platforms trust the agent, then check the damage. Heimdall decides what the agent can do before its container starts. The container never holds the key.
Git caught code collisions. Humans caught design collisions through communication. AI agents have neither. Shepherd detects when individually correct agents build a collectively broken system.
Production-grade guardrails for AI agents: auditability, real-time safety, and self-healing recovery without changing how you build.
KalaaMitra is an autonomous multi-agent co-pilot that eliminates daily friction for CBSE primary Art & Crafts teachers in no-smartboard classrooms.
Kiosk managers describe faults in plain language. Site Agents search exact-compatible verified playbooks first, and hand unresolved cases to developers as structured Evidence Packs.
LastTube gives every discontinued-shade candidate a system or human disposition, then unlocks an observed offer only when exact-variant and visual proof are complete.
InvoiceLoop turns AI-extracted invoice data into decisions people can review, approve, and stand behind.
An advanced, offline-first Autonomous AI Code Writer and Multi-Agent Operating System powered entirely by local models.
CinePilot is your AI production crew—turning creative ideas into production-ready plans with intelligent budget, schedule, risk, and research insights.
Clanker-in-the-middle: Middleware that watches, gates, and logs every move your AI coding agent makes.
Giral asks Gemini the questions your buyers ask, tracks if your brand gets cited over competitors, diagnoses why it doesn't, and turns that into a prioritized fix-it list — for AI-search visibility.
State-bound release control for autonomous agents: certify candidates by diffing real effects, then gate promotion with an audited, tamper-evident ledger.
Users inspect a synthetic factory, optimize an isolated scenario, compare it to the committed baseline, and request human approval.
Agent middleware that runs a plan's tool calls before the model has finished writing the plan, and refuses the ones it cannot prove safe. 2.19x faster end to end
