Experiments & open ideas
Our running backlog of 20+ experiments — agent tooling, security, filesystems and more. Some already have people building on them; many have no contributors yet — you could be the first. Say hi on Discord.
MCP proxy or sub-agents to offload agent memory, shrink the main context and cut tokens.
A structured prompt that lets AI agents resolve merge conflicts intelligently.
MCP as a REST replacement
agentsShowcase how MCP tools outperform REST endpoints for agent interactions.
Rerun.io MCP server
agentsExpose Rerun's 3D visualization APIs as an MCP toolset for agents.
Reset passwords with LLMs via a semantic ZK-like proof.
Vibe-coding context injection
agentsInject style or mood context into coding agents to influence output tone.
Compact an agent context window
agentsAuto-summarize and compress code state for long agent sessions.
Ad-hoc UI in chat agents
agentsLet agents spawn temporary input forms dynamically inside chats.
Showcase n8n
toolingDemonstrate n8n workflows powered by LLM actions and triggers.
Malware-detector AI agent
securityAn agent that scans and flags malicious code.
Sommelier assistant with skills
aiRecommend wines using AI skills for pairing, taste and region.
git-clone post-op scanner
securityScan newly cloned repos via an automated AI prompt workflow.
LLM linter
toolingEnforce linting rules and propose fixes using AI coding agents.
Use error-correcting codes to recover data in rencfs.
Use a passkey as a second factor to unlock the encrypted filesystem.
Add an IPFS storage backend to the encrypted filesystem.
Refactor rencfs examples into standalone crates.