Technology Stack

Model Agnostic. Architecture First.

We select technology based on the requirements of each system — not allegiance to a single provider. The right tool for the right job, integrated into a coherent architecture.

Technologies we work with — not partners: The tools listed here are technologies Axioprax uses in system design and engineering. We do not claim official partnerships, reseller status or endorsed relationships with any of these providers unless explicitly stated.

AI Models

We are model-agnostic. Selection depends on the task, cost profile and data privacy requirements of each system.

Anthropic Claude

Strong reasoning, long context handling and instruction following. Used for complex agent tasks and document intelligence.

OpenAI GPT / o-series

Broad capability, strong function calling and tool use. Used for classification, extraction and agentic task execution.

Google Gemini

Multimodal capability and Google ecosystem integration. Used where vision inputs or Google Workspace context matters.

Open-source models

Locally deployed or self-hosted models where data cannot leave the customer environment.

Engineering

The languages, frameworks and AI-assisted development tools we build with.

Python

Primary language for AI/ML pipelines, agent systems, data processing and backend services.

TypeScript / Node.js

Used for API layers, Next.js applications and web-facing AI interfaces.

React / Next.js

Frontend framework for AI-native web applications and internal tools.

Claude Code

AI engineering environment used to accelerate development velocity across client and lab projects.

Automation & Orchestration

Platforms and tools for workflow automation and agent orchestration.

n8n

Open-source workflow automation. Used for connecting business systems, webhook handling and automation pipelines.

LangGraph

Graph-based agent orchestration for complex multi-step agent workflows with explicit state management.

Make (Integromat)

Low-code automation for specific integration scenarios and client environments where n8n is not preferred.

Custom Orchestration

Bespoke orchestration layers built in Python or TypeScript when standard platforms do not fit the architecture.

Agent Infrastructure

Protocols, frameworks and infrastructure for building production-grade AI agent systems.

MCP (Model Context Protocol)

Anthropic's open protocol for connecting AI models to tools and data sources. We build and consume MCP servers.

OpenAI Assistants API

Used where thread management, file handling and tool orchestration within the OpenAI ecosystem are required.

Vector Databases

Pinecone, pgvector (PostgreSQL) and Weaviate for semantic search and RAG retrieval infrastructure.

Agent Observability

Custom trace logging, token tracking and cost attribution for AI systems operating in production.

Enterprise Systems

Business platforms we integrate AI systems with across sales, support, HR and operations.

Salesforce

CRM integration for lead enrichment, opportunity updates, activity logging and pipeline reporting.

HubSpot

CRM and marketing integration for sales operations automation and contact management.

Google Workspace

Drive, Gmail, Calendar and Docs integration for knowledge management and executive automation.

Microsoft 365

SharePoint, Teams and Outlook integration for enterprise knowledge systems and operational workflows.

Zendesk / Freshdesk

Customer support platform integration for ticket classification, knowledge retrieval and response automation.

Jira / Linear

Engineering workflow integration for sprint planning assistants and incident management automation.

Data

Data storage, processing and retrieval technologies used in AI system architecture.

PostgreSQL

Primary relational database. Used with pgvector extension for combined structured storage and vector retrieval.

Redis

Caching, session management and lightweight queue infrastructure for high-throughput AI pipelines.

Document stores

MongoDB and similar for document-oriented data where flexible schemas suit the use case.

Cloud & Infrastructure

Cloud platforms and deployment infrastructure for AI systems.

AWS

Primary cloud platform for many client deployments. EC2, Lambda, S3, RDS and Bedrock.

Google Cloud

Used where Google Workspace integration, Vertex AI or BigQuery are central to the architecture.

Azure

Microsoft ecosystem deployments, Azure OpenAI Service and Active Directory integration.

Our Approach

Technology Follows Architecture

We start with the problem and the system requirements — then select technology. We do not fit problems to a preferred stack. This means we can integrate into client environments rather than requiring migration to ours.

No lock-in to a single model provider

We design systems to swap models as the landscape evolves. Prompt logic is separated from model configuration.

Integrates with your existing stack

We build to your tools, CRM, data platform and cloud. Replacement is never a prerequisite for adding AI.

Production-grade from the start

Error handling, observability, access control and cost management are built in — not added later.

Technical Discussion

Want to Discuss the Architecture?

Tell us about your existing stack, the systems AI needs to connect with and the constraints of your environment. We will tell you how to approach the architecture.