The field-tested toolbox behind R21 Digital & my open-source work
The models, agent/MCP infrastructure, retrieval, voice, and app stack I actually build and ship production systems with.
Not a link dump — every tool here I've shipped with, run in production, or evaluated hands-on for a real project. ★ = day-to-day core.
| Tool | What I use it for | |
|---|---|---|
| ★ | Anthropic / Claude | Primary model family for agentic + tool-calling work. |
| ★ | OpenRouter | Single gateway for multi-model routing (cost/latency per task). |
| ★ | Groq | Low-latency inference for fast agent steps. |
| OpenAI | GPT models via the API. | |
| Google AI Studio / Gemini · Vertex AI | Gemini models, prompt tooling, GCP-hosted inference. | |
| Mistral AI | Open-weight models. | |
| Moonshot / Kimi | Long-context model API. | |
| xAI / Grok | Grok model console. | |
| Hugging Face | Model hub, hosted inference, Spaces. | |
| NVIDIA NIM | GPU-served model endpoints. |
| Tool | What I use it for | |
|---|---|---|
| ★ | Model Context Protocol | The standard I write servers/tools against — my published server. |
| ★ | Make.com | Production automation orchestration for client work. |
| ★ | n8n + n8n-mcp | Self-hostable workflow automation + AI nodes. |
| Relevance AI | Agent + tool marketplace platform. | |
| Rasa | Open-source conversational NLU / dialogue framework. | |
| Huginn | Self-hosted "agents that watch and act." | |
| Context7 · Playwright MCP | Live library docs and browser control, as MCP tools. | |
| ClickUp · Metricool · NotebookLM | Business systems wired to agents over MCP rather than scraped. |
| Tool | What I use it for | |
|---|---|---|
| ★ | Pinecone | Vector DB for RAG + agent memory. |
| ★ | Exa | Neural search API for agents. |
| ★ | Tavily | Search/extract API purpose-built for LLMs. |
| ★ | Firecrawl | Turn sites into LLM-ready markdown. |
| LangSmith / LangChain | Tracing + eval for LLM apps. | |
| Brave Search API | Privacy-first web search. | |
| SerpApi | Structured search-engine results. | |
| Apify | Full-stack scraping + data extraction. | |
| SearXNG | Self-hosted metasearch. |
| Tool | What I use it for | |
|---|---|---|
| ★ | ElevenLabs | TTS / voice generation. |
| ★ | Deepgram | Real-time speech-to-text. |
| ★ | faster-whisper | Local, private transcription — my default (no paid API). |
| AssemblyAI | Transcription + audio intelligence. | |
| Speechmatics | Multilingual STT. | |
| Vapi · Retell AI | Voice-agent frameworks. |
| Tool | What I use it for | |
|---|---|---|
| ★ | Supabase | Postgres + auth + storage for most products. |
| ★ | PostgreSQL | Primary relational store. |
| ★ | Vercel | Deploy target for Next.js apps. |
| ★ | AWS | SES email, EC2/PM2 hosting, HIPAA-eligible infra. |
| Redis | Cache / queues / rate-limiting. | |
| Railway · DigitalOcean | Container / VPS hosting. | |
| Clerk | Auth + user management. | |
| Firebase | Realtime / auth where it fits. |
| Tool | What I use it for |
|---|---|
| ★ AWS SES | Transactional email standard across R21. |
| Postmark · SendGrid | Transactional email. |
| WAHA | Self-hosted WhatsApp HTTP API. |
| AgentMail | Inboxes for agents. |
| Tool | What I use it for | |
|---|---|---|
| ★ | Next.js · React · Tailwind | Core web stack. |
| ★ | v0 by Vercel | AI UI generation. |
| Figma | Design + handoff. | |
| 21st.dev | Component library / inspiration. | |
| Retool · Appsmith | Internal tools & admin panels. | |
| WeWeb · Firebase Studio · Emergent | No/low-code builders (evaluated). |
| Tool | What I use it for | |
|---|---|---|
| ★ | Flutter / Dart | One codebase to six targets. Shipped a tournament + catch-logging app that has been in the field since 2025. |
| ★ | PWA | For when a store listing is the wrong answer. I write the service worker by hand because the caching rules are specific to the app and a plugin buries them. |
| App Store Connect · Play Console | Store submission, review policy, account-deletion requirements. |
| Tool | What I use it for | |
|---|---|---|
| ★ | Claude Code | Primary agentic dev environment. |
| ★ | OpenAI Codex | Second-model adversarial review before anything public ships. It catches what Claude and I both missed. |
| t3code | Multi-agent orchestration over one repo. I contribute upstream to it. | |
| Cursor | Editor-native agent work. | |
| Warp | Agentic terminal. | |
| Windsurf · Kiro · Factory · Qoder · Cosine · Kilo Code | Evaluated AI IDEs / agents. | |
| Replit · Bolt.new · Blink | Prototyping sandboxes. |
| Tool | What I use it for |
|---|---|
| Runway · Remotion · Shotstack | Video generation / programmatic video. |
| Midjourney · Stability AI · DALL·E | Image generation. |
| Kie.ai | Aggregated media-generation APIs. |
| Tool | What it is |
|---|---|
| Sentry | Error tracking + performance. |
| Metabase | BI / dashboards. |
| Cal.com | Scheduling infrastructure. |
| Chatwoot | Omni-channel support desk. |
| Mautic · EspoCRM · OpenProject | Open-source marketing / CRM / PM. |
| Tool | What I use it for |
|---|---|
| ★ PostHog | Product analytics — funnels on the paths that actually matter, self-hostable if the data is sensitive. |
| Stripe | Payments & billing. |
| RevenueCat | In-app subscriptions. |
| Mapbox | Maps / geodata. |
| Twelve Data | Market data API. |
| Apollo | B2B data + enrichment. |
| PageSpeed Insights · Search Console | Performance + GEO/SEO. |
Maintained by Carlos Jimenez — Forward-Deployed / Applied AI Engineer.
Curated from a working practitioner's toolbox; opinions are field notes, not endorsements.
Curated from a working practitioner's toolbox; opinions are field notes, not endorsements.