The LLM market changes every week. Cheaper models emerge, others get discontinued. Why lock yourself into one expensive API? MyDatAgent's AI Gateway is the intelligent layer between your code and models — one key, all providers, total governance.
1from openai import OpenAI 2 3client = OpenAI( 4 api_key="mda-prod-7c9f...8e21", # <-- your MDA key 5 base_url="https://gateway.mydatagent.ai/v1" # <-- only change the URL 6) 7 8resp = client.chat.completions.create( 9 model="mda/llm-2.1-32b", # or "openai/gpt-4o", "anthropic/claude-opus" 10 messages=[{"role": "user", "content": "Refactor this function."}], 11) 12 13print(resp.choices[0].message.content)
1import OpenAI from "openai"; 2 3const client = new OpenAI({ 4 apiKey: process.env.MDA_KEY, // <-- your MDA key 5 baseURL: "https://gateway.mydatagent.ai/v1" // <-- only change the URL 6}); 7 8const resp = await client.chat.completions.create({ 9 model: "mda/llm-2.1-32b", 10 messages: [{ role: "user", content: "Summarize the report." }] 11}); 12 13console.log(resp.choices[0].message.content);
1curl https://gateway.mydatagent.ai/v1/chat/completions \ 2 -H "Authorization: Bearer $MDA_KEY" \ # <-- your MDA key 3 -H "Content-Type: application/json" \ 4 -d '{ 5 "model": "mda/llm-2.1-32b", 6 "messages": [{"role":"user","content":"Hello."}] 7 }'
OpenAI-compatible reverse proxy architecture. Your app becomes completely vendor-agnostic — routing changes in the dashboard, not in code.
You use the OpenAI SDK on your backend, but point to MyDatAgent's URL. Code stays the same. Models change in the dashboard.
mda/llm-2.1-32b for anthropic/claude-opus and done.The AI Gateway enables dynamic substitution. Configure routes so high-cost requests are automatically redirected to MyDatAgent's SLMs running on private infrastructure.
Devs love using Claude Code, but Opus's per-token cost destroys the IT budget. Code review, simple refactoring, tests — everything at premium prices.
Dev still uses Claude Code. But the Gateway intercepts and routes: simple task goes to private MDA LLM 2.1 32B; only critical refactoring hits Opus.
That dev who pastes the company key on their laptop? That analyst who uploads a script with the server's key? Over. User-level governance with quotas, rate limits, and auditable logs.
The company's main credential never leaves the MDA backend. Devs and analysts get only derived, traceable, revocable keys.
Max spend (USD) per user/month. Tokens per minute (TPM). Requests per minute (RPM). When the budget runs out, the key stops consuming.
Who · when · which model · which tokens · what it cost · which project. Auditable, exportable logs for SIEM. Your CISO sleeps soundly.
Primary model down? Gateway redirects to secondary automatically. Built-in resilience, no extra code in your app.
From immediate URL swap to full agent orchestration — pick the integration that fits your current architecture.
Swap OpenAI's base URL in your code for MyDatAgent's. Done. You're already running our high-performance SLMs, in USD, with zero FX risk, and ultra-low latency from our Brazilian datacenter.
Point your app to the Gateway and let MyDatAgent's intelligence route dynamically. Everything logged, everything governed, single endpoint.
Don't just consume tokens? Use the API to orchestrate MyDatAgent's ready-made agents (SDR, BI, DataOps) right in your system — Webhooks, n8n, or REST with MCP.
Migrating from Big Tech APIs to private SLMs requires expertise in prompt routing, quantization (FP8), and RAG fine-tuning — so you don't lose quality.
MDA Consulting does the migration for you. We audit your current API logs, design the routing architecture (what goes to SLM, what stays on GPT-4), set up the Gateway with key governance, and guarantee your savings in the first 30 days.