Agents fail on the harness, not the model.
Agent operating systems, Solana infrastructure, open source.
Personal brand / hub: nyk.dev · GitHub: github.com/0xNyk
Ventures under the name: co-founder rpc edge · founded Builderz · selective consulting.
I build the ops layer for AI agent fleets - quotas, safety gates, orchestration, control planes - and prove it in daily use: my ventures run on the same agent stack I open-source, down to the live star counts on this page. 16.3K+ OSS stars (16,268 live across every repo I have authored, listed on this page), code upstream in Hermes Agent. rpc edge is the Solana infrastructure venture; the rest is public tooling people use.
rpc edge - low-latency Solana infrastructure for high-frequency trading. Dedicated RPC, Yellowstone gRPC, decoded shred streams, and a transaction sender - co-located with the cluster and the Jito Block Engine. Metered bandwidth, self-serve, settled in USDC on Solana. By Polaris Labs.
Open tooling for the same stack:
| Repo | What it is |
|---|---|
| rpcedge-toolkit | TypeScript SDK, CLI, and MCP server for Solana trading infrastructure |
| rpcedge-copy-ref | Paper copy-watch reference - doctor → poll wallet → paper log (not financial advice) |
| rpcedge-dlmm-ref | Paper Meteora DLMM pool-watch reference - doctor → accountSubscribe → paper log (not financial advice) |
→ rpcedge.com · docs.rpcedge.com · toolkit
Latest open source (shipping):
| Project | One-liner | Stars |
|---|---|---|
| rpcedge-toolkit | TypeScript SDK, CLI, and MCP server for rpc edge / Solana trading infra |  |
| rpcedge-copy-ref | Minimal paper copy-watch reference for rpc edge (doctor → poll → paper log) |  |
| rpcedge-dlmm-ref | Paper Meteora DLMM pool-watch reference for rpc edge (not financial advice) |  |
| hermes-buzz | Hermes Agent × Buzz starter kit (buzz-acp + hermes-acp onboarding) |  |
| llmquota | Terminal roster for LLM CLI quotas - see usage, reset times, and which agent still has headroom |  |
| agent-security | Safety gates for agent-touched repos - scan leaks, vet code, trip on prompt injection, guard destructive ops |  |
| silo | Isolated Claude Code profiles via CLAUDE_CONFIG_DIR - no credential vault swap |
 |
| unmachined | Anti-AI-slop skill - text + UI that reads written, not generated |  |
Upstream: contributor to Hermes Agent (Nous Research) - cron/timezone fix on main via co-authored commit, plus open PRs for session continuity, skills, and subagent isolation. Also proposing fixes to OpenClaw.
Fixed scope, fixed price band, agreed before anything starts. You get the system and the handoff, not a dependency on me.
| Agent Reliability Audit | 2 weeks · from $4K | Trajectory evals that locate where your agents actually fail, plus a fix list in priority order |
| Agent OS Setup | 2-4 weeks · from $5K | A working agent stack your team can run without me |
| Solana infra | 2-6 weeks · from $8K | Latency-sensitive execution paths, data and transaction sending |
| Agent OS Kit | €149 · instant | The controls as editable files, if you would rather install it yourself |
→ See all packages · Book a 30-min intro
Top 10 by stars. Live badges. Profile: github.com/0xNyk.
| # | Project | What it does | Stars |
|---|---|---|---|
| 1 | mission-control | Agent fleet dashboard - tasks, quality gates, cost tracking, real-time orchestration |  |
| 2 | awesome-hermes-agent | Curated skills, tools, and integrations for the Hermes Agent ecosystem (Nous Research) |  |
| 3 | council-of-high-intelligence | Multi-perspective reasoning skill - historical thinkers deliberate on your problem |  |
| 4 | marketing-dashboard | Marketing ops control center for AI agent teams - CRM, outreach, content, analytics |  |
| 5 | lacp | Local Agent Control Plane - policy gates, memory layers, hook pipeline |  |
| 6 | xint | X intelligence CLI (TypeScript + Bun) - search, monitor, analyze, engage |  |
| 7 | awesome-agent-cortex | Sovereign agent stack - scripts, on-chain identity, knowledge graphs |  |
| 8 | builderz-solana-dapp-scaffold | Production-ready Solana dApp starter (Next.js, Tailwind, web3.js) |  |
| 9 | openclaw-to-hermes | Battle-tested migration from OpenClaw to Hermes Agent |  |
| 10 | xint-rs | X intelligence CLI as a single Rust binary - <5ms startup |  |
More open-source repositories - sorted by stars
| Project | What it does | Stars |
|---|---|---|
| builderz-xNFT-scaffold-next | Solana xNFT scaffold (Next.js, TypeScript, Tailwind) |  |
| solana-claude-md | Open-source CLAUDE.md pack for Solana program work |  |
| hermes-cf-bypass | Cloudflare TLS fingerprint bypass for Hermes on datacenter VPS |  |
| unmachined | Anti-AI-slop agent skill - deterministic scanners + severity-tiered tell catalogs for text and UI |  |
| renaissance-xnft | Royalty tracking and redemption for NFT communities |  |
| silo | Isolated Claude Code profiles via CLAUDE_CONFIG_DIR - personal / work / clients without vault swap |
 |
| obsidian-curator | Deterministic knowledge-graph maintenance for Claude + Obsidian |  |
| homebrew-xint | Homebrew tap for xint and xint-rs |  |
- Infrastructure wins on latency, reliability, and sharp defaults.
- Agent teams need operations, not demos.
- The strongest stack is open-core control plane + premium operator UX.
| Profile | github.com/0xNyk · nyk.dev · @nykdotdev |
| rpc edge | rpcedge.com · @rpcedge · docs |
| Studio | Builderz · consulting |