AI Engineer · Full Stack Architect · Founder @ NorahLabs

Building production AI systems from Nairobi, Kenya. Agentic orchestration engines, RAG pipelines, and native desktop applications — designed for developers who want real architectural understanding without cloud lock-in.

A native desktop AI coding environment for developers working on non-trivial codebases.

v8.7.5 · Mac + Linux · atlarix.dev

Atlarix builds a section-scoped Blueprint graph on demand using a four-layer index stack — Universal Ctags (symbol index), ast-grep (import/call/HTTP route edges), BM25 (semantic symbol ranking), and ripgrep (text fallback) — delivering ~6,500 tokens of structural understanding for a 99-file TypeScript section in a 25-section multi-repository workspace.

Published research: Blueprint: Section-Scoped Structural Graph Retrieval and Post-Turn Compression for Agentic LLM Coding in Multi-Repository Workspaces · Zenodo (CERN), May 2026

v8.7.x highlights:

• Section-scoped get_blueprint — focus_path required; preflight path validation; richness-gated hybrid merge (ctags + ast-grep → hybrid when thin)
• Blast radius BFS over Blueprint flow edges before every write/edit approval
• Inline Mermaid diagrams generated on demand from codebase scan — no stale DB snapshot
• Post-turn tool-result summarisation — 95–98% per-read compression before history persistence

v8.6.x highlights:

• Universal Ctags + ast-grep Blueprint stack — replaced previous parser approach
• Route edges: Express/Fastify/Next.js HTTP handlers as first-class structural edges
• ast-grep worker pool with Python/Go dynamic grammar registration
• Incremental file-watcher refresh (500ms debounce) — per-file ctags + ast-grep, not full rebuild

v8.x foundations:

• ~50% pre-dispatch token reduction — 6 auto-injected blocks converted to tool-driven retrieval
• compress_context — model-triggered, not pre-dispatch
• Tool consolidation 60+ → 28 — unified search (auto/text/semantic-graph/structural), edit_file, on-demand MCP, Skills via ToolRegistry
• Skills Marketplace — community registry at atlarix-skills

v7.x foundations:

• Parallel agents: Research · Architect · Builder · Reviewer · Debugger · Explore
• DelegationMonitor — cyclic and runaway delegation detection
• Compass managed tier — Fast / Balanced / Thinking, no API key required
• Full CI/CD: Apple Notarisation + Linux packaging (.deb / .rpm / .AppImage)
• Three-tier subscription: Free · Pro ($19/mo) · Workforce ($79/mo)

Multi-provider: OpenAI · Anthropic · Google Gemini · Groq · Mistral · xAI · OpenRouter · Together AI · Ollama · LM Studio

Achievements: Winner — Amazon Nova AI Hackathon (Devpost × AWS, 2026) · GDG Nairobi Agentathon (May 2026) · Lua x Antler (Nairobi, 2026) · Red Bull Basement Global Innovation Programme

Blueprint: Section-Scoped Structural Graph Retrieval and Post-Turn Compression for Agentic LLM Coding in Multi-Repository Workspaces Amariah Kamau, NorahLabs — May 2026

Documents the Blueprint architecture, post-turn tool-result summarisation, and benchmark results from a controlled A/B evaluation on a production multi-repository workspace. Honest findings — including a counterintuitive result about total context usage.

atlarix-skills — Community skill registry. SKILL.md files teaching agents language patterns and framework conventions across React, Next.js, Python, TypeScript, Go, Rust, Docker, MCP, and more. Apache 2.0.

atlarix-mcps — MCP server registry for Atlarix. One-click connections to Gmail, Calendar, Drive, and more from within the agent environment. Apache 2.0.

atlarix-releases — Desktop app release builds for Mac and Linux.

Praxia — AI healthcare assistant with MONAI X-ray analysis (pneumonia, fractures, tumours), multilingual symptom diagnosis, real-time WebSocket chat, and Docker/Nginx/Celery production stack.

Commit Checker — Open source CLI for Git commit analysis.

Atlarix is built in Nairobi. Most serious AI developer tooling assumes a Western cloud subscription. Atlarix is open-model by design — bring any provider, including locally-built African models. Currently piloting with African AI labs for native MansaLLM integration.