Curated Open-Source AI Repositories

“I red-team every MCP server with JWT + OPA + HITL before it touches Razorpay — pentagi automates that adversarial loop as a swarm instead of a checklist. Its autonomous pentest pattern is the pre-deploy gate I am adding to my Junagadh ship pipeline for Gujarat fintech clients.”

“I built Curro as an AI content studio that writes like its owner — openhuman pushes the same idea further with local-first life memory plus fleet orchestration in Rust. Its memory-of-your-life architecture is the direction I am taking Curro for Junagadh founders who refuse cloud journals.”

“I track career-ops (batch 9) where local-first job search scored listings A-F — ai-job-search adds eval-driven CV tailoring inside Claude Code, the same evaluate-before-act loop in my QA swarms. Its runs-on-your-machine guarantee is what I prescribe to Junagadh devs targeting the 117K open AI roles in India Sep 2026.”

“I run OpenClaw-style assistants for Gujarat SME clients where agents need memory plus sandboxed tools without a Python dependency mess. OpenFang ships one 32MB Rust binary with cold start under 200ms from my Junagadh lab tests, and its WASM dual-metered sandbox is the pre-deploy gate I enforce before any agent touches Razorpay keys.”

“I test minimal harnesses on a Rs 6K VPS where every extra dependency breaks deploys. GenericAgent runs under 30K context where others burn 200K-1M, and its crystallize-each-task-into-skill loop matches how I grow my 41-skill Junagadh catalog after every client rollout.”

“I track token burn per agent run in my P95 ledger where verbose tool outputs eat margin on fixed-price Gujarat SME builds. Caveman cut my Claude Code review loops by roughly half in Junagadh trials, and I keep it as a proxy gate in front of expensive reasoning calls on the VPS.”

“I ship browser-use automation for Rajkot RFQ inboxes where agents stall on logins and Cloudflare. Agent S3 beating human OSWorld at 72.6 percent is the first computer-use result I trust for unattended runs, and I study its ACI grounding loop from Junagadh before wiring any GUI step into client pipelines.”

“I benchmark 70B offline on laptop where long-horizon repo sessions die from context loss. Kimi K3 holding 1M context with native vision for CAD plus kernel work is the open-weight I point Surat clients to for overnight agent runs, and I pair it with my trace-MCP harness to keep tool calls inside budget.”

“I roll out shared agent setups for 5-10 staff Gujarat SME teams where everyone hand-rolls prompts and context dies per seat. TencentDB extracting reusable Skills plus CodeGraph from past sessions is the cold-start fix I prescribe from Junagadh, and I test its proxy mode on the Rs 6K VPS with zero-code client switching.”

“I scan AI news nightly for my journal from Junagadh, and hand-rolled aggregators rot within weeks. WorldMonitor fusing news plus geopolitical plus infra signals into one dashboard is the situational layer I want above my SearXNG scanner. I would self-host it on the Rs 6K VPS and point my research agents at its feed before any manual browsing.”

“I run LangGraph plus CrewAI-pattern swarms behind one OPA gate in Junagadh, and cross-harness coordination is still manual glue. Ruflo as a meta-harness with adaptive memory plus RAG that speaks Claude Code, Codex and Hermes natively is the consolidation layer my multi-harness setup is missing. I would trial it against my LangGraph checkpoint rig on catalog QA first.”

“I work with Gujarat retail shops where cameras trigger DPDP anxiety but footfall data drives staffing. RuView doing presence plus vital signs from WiFi with no video is the privacy-shaped sensor I would actually get permission to install. I would pair it with my WhatsApp footfall alerts on the Rs 6K VPS and keep every byte inside the shop network.”

“I measured 42 minutes of agent exploration per unfamiliar repo in my Junagadh lab before codebase-memory-mcp cut it to milliseconds. Code-review-graph doing the same local-first map with benchmarked context reductions, purpose-built for reviews, is the exact pre-filter I bolt in front of expensive reasoning calls. I would wire its MCP server into Cursor morning scans beside trace-mcp.”

“I run client workloads on a Rs 6K VPS from Junagadh where every deploy is SSH plus memory. 1Panel giving that box a panel plus AI management without cloud rent is the ops layer my Gujarat SME fleet wants. I would trial it beside my current hardened Ubuntu image and keep whichever survives a monsoon patch cycle.”

“I ran paper-trading research summaries for a Surat client with TradingAgents capped at end-of-day reports. Kronos as a market-native foundation model is the forecasting backbone I would eval next for that desk — language of markets instead of general chat bolted onto prices. I would keep it read-only research, never live execution, same gate as before.”

“Reliable citations for India compliance (DPDP/GST) — grounds Hindi/Gujarati docs before answering, vs hallucinated RAG.”