API Design & Development
We don't just design APIs — we run them. CogniGraph serves graph queries over HTTP. IronFlow exposes workflow execution via REST. PLU Finder's API handles thousands of requests daily. When we design an API for your project, the patterns come from production experience, not a textbook.
APIs We Run in Production
Our own products prove our API design — every day, under real load.
CogniGraph API
In-house — Knowledge Graph over HTTP
Full CRUD for documents, typed graph relationships, and four search modes (vector, semantic, hybrid, graph-augmented) — all via clean Axum endpoints. Lua scripting execution endpoint for server-side graph operations. Pluggable backends behind a single API surface.
IronFlow API
In-house — Workflow Execution over REST
Submit Lua workflows via POST, get structured results back. Run history, validation endpoints, node listing, webhook routes with auth. Designed for embedding workflow automation into any service via HTTP.
PLU Finder API
In-house — Produce Intelligence
Our produce-data API provides lookups, search and category access for applications and integrations. Its production use informs our approach to stable contracts, access plans, bounded queries and measured performance.
MCP Servers
In-house — AI Tool Integration Protocol
MCP servers expose scoped business tools and data to AI applications, including PLU Finder. We design schemas, authentication, authorization and error handling together. A callable tool is an integration surface, not permission for a model to perform any operation it requests.
API Paradigms
We pick the right style for the problem — not the one we used last time.
REST
Our default for most projects
- OpenAPI 3.1 specification-first design
- JSON:API standardized response formats
- Versioning via URL, header, or content negotiation
- HATEOAS for discoverable APIs
GraphQL
When clients need flexible queries
- Type-safe, self-documenting schema design
- Subscriptions for real-time data via WebSocket
- Federation for distributed graph architecture
- Efficient resolver strategies to avoid N+1
tRPC
End-to-end TypeScript type safety
- Types flow from backend to frontend automatically
- Catch API errors at compile time, not runtime
- RPC-style procedure design
- React Query integration for seamless data fetching
Real-Time
WebSockets and event-driven
- WebSockets for bidirectional communication
- Server-Sent Events for one-way streaming
- Pub/Sub patterns for event-driven systems
- Used in IronCrew's agent messaging and streaming
Where We Deploy APIs
From database-embedded microservices to global edge — we match deployment to requirements.
ArangoDB Foxx
APIs inside the database
Database-local execution reduces application-to-database round trips. JavaScript/TypeScript with Express-like routing, graph traversals in single requests, ACID transactions. We've built Foxx services for GameStake, IoT platforms, and enterprise integrations.
Application Servers
Axum, Fastify, FastAPI, OpenResty, RESTEasy
Axum (Rust) for our high-performance tools — CogniGraph, IronFlow, IronCrew. Fastify and Hono (Bun/Node) for web APIs. FastAPI (Python) for data-heavy services. OpenResty (Lua) for API gateways. RESTEasy (Java) for enterprise microservices.
Edge & Serverless
Cloudflare Workers, Netlify Functions, AWS Lambda
Cloudflare Workers, AWS Lambda and Netlify Functions for workloads suited to managed functions. We evaluate execution limits, persistence, identity integration, latency and cost before choosing a target; application requirements determine the deployment.
Platform Hosting
Railway, Docker, Kubernetes
Railway for fast service deployment with automatic scaling and managed databases. Docker and Kubernetes for containerized deployments on any cloud. Our Rust tools also ship as single binaries — no container required for air-gapped or edge environments.
Security & Quality
Built in from the start, not bolted on after launch.
Authentication & Security
- OAuth 2.0 / OIDC and scoped service credentials
- Server-side authorization for each resource and operation
- Validated inputs, bounded payloads and rate limits
- Idempotent mutation design and explicit approval where required
Documentation & DX
- OpenAPI specifications, examples and client integration guidance
- Contract tests for successful and failed requests
- Documented pagination, versioning and error semantics
- Streaming lifecycle, cancellation and retry behavior
Proven Track Record
APIs we've owned, operated, and evolved over years.
vzaar
7+ years — Video Hosting API Platform
Complete API platform ownership — client libraries in PHP, Java, ActionScript 3, C#. Hundreds of third-party integrations. Developer documentation, SDKs, and continuous API evolution serving a global video hosting customer base.
GameStake Technologies
Gaming Platform API Architecture
ArangoDB Foxx Microservices for game state and player data. OpenResty replacing Nakama Server for custom game logic. Payment integration (Stripe, Skrill). Real-time gaming events via WebSocket.
PLU Finder
In-house — Production API at Scale
api.plufinder.com and mcp.plufinder.com serving thousands of monthly users across 4 domains. 10-tier subscription model. REST API for produce data plus MCP server for AI application integration.
CogniGraph & IronFlow
In-house — Rust HTTP APIs
Both ship with built-in REST APIs — document CRUD, search endpoints, workflow execution, validation, run history. Designed API-first so they can be embedded in any architecture via HTTP.
Ready to build APIs that last?
Contact us to discuss your API design — we'll bring patterns from APIs we run in production, not from a framework tutorial.
Get in Touch