AI integration
AI agents, RAG and MCP integrations built into real products, using Claude, OpenAI and Gemini.
What we build
We add AI to products that already exist and build AI-first tools from scratch. The team has shipped LLM features behind a multi-provider layer covering OpenAI, Anthropic Claude and Google Gemini, retrieval-augmented generation on PostgreSQL with pgvector, custom agents for internal workflows, and MCP integrations.
Our own open-source work includes a Claude Code plugin that produces narrated videos and an on-device speech dictation app. Ariful’s work at weDevs includes an AI form builder and AI generators inside WordPress plugins with 35,000+ active installs.
Typical deliverables
- LLM features in your product with provider fallback and cost controls.
- A RAG pipeline: embeddings, a vector index, retrieval and grounded answers.
- Agents and MCP tools that connect a model to your systems, with human approval where mistakes would be costly.
- Evaluation of outputs against examples you provide.
- Documentation of prompts, limits and how to change providers.
How we work
We begin with the task you want done and a handful of real examples, then prototype quickly to find out what is feasible before committing to a full build. We are open about what models do poorly and design around it.
Who works on this
Engineers from the core team who have shipped this kind of work. Their public profiles show the details.
Related work
Software the team built and released in the open.
- DevWizard productmacOS 14+
HoldTalk
Free, offline push-to-talk dictation for macOS. Hold a key, speak Bangla or English, release, and the text appears at your cursor.
Versioned releases on GitHub, installable with a single checksum-verified script.
- Swift
- whisper.cpp
- Whisper large-v3-turbo
- AVFoundation
- DevWizard productClaude Code
Claude Video Studio
A Claude Code plugin that acts as a video production agency, turning a real web app into a reviewed, narrated, branded tutorial or marketing video.
- TypeScript
- Playwright
- Remotion
- ffmpeg
Questions before you write to us.
Short answers on how we work. Anything else, ask by email.
How do you handle our data in AI features?
We agree what data may leave your systems before building, keep secrets out of prompts and logs, and document what each provider receives. An NDA can be signed first.
Are you tied to one AI provider?
No. We build behind a provider layer covering OpenAI, Anthropic Claude and Google Gemini, so you can switch models or add a fallback without rewriting the feature.
Planning ai integration?
Tell us the goal and the constraints. We reply with questions, then a written scope and quote.