Give your product a brain it owns.
Drop-in agentic chat with a living knowledge base. OpenAI-compatible, streaming, multi-tenant — built on local models you control.
Chat with Toby.
Toby is running on the Atlas API right now, answering from a sandboxed knowledge base. Every keystroke hits a real endpoint.
Everything you'd build yourself.
Already built.
Atlas gives you the tooling of a frontier lab with the latency of a local server.
Agentic Chat
OpenAI-compatible completions endpoint. Streaming, sessions, tool use. Point your existing client at Atlas and it just works.
Living Knowledge Base
Upload PDF, DOCX, HTML, MD, TXT, JSON, CSV. Atlas chunks, embeds, and searches it — with citations baked into every answer.
Multi-Tenant
Each project gets an isolated knowledge base, sessions, and API keys. Build a feature today, spin up a whole new product tomorrow.
Streaming by Default
Server-sent events. Your UI gets tokens the moment they're generated, with tool-call events surfaced as separate streams.
Local Inference
Runs on your own hardware (or ours). No per-token fees. No data leaves the cluster. Load-balanced across backends for concurrent users.
Built for Humans
Personas that persist per-tenant. Change the name, tone, or rules with a single PATCH request. Live reload, no redeploys.
Five lines of code.
If your stack already speaks OpenAI, you already speak Atlas.
// Point any OpenAI client at Atlas.
const url = "https://atlas.moliam.com"
+ "/v1/chat/completions";
const res = await fetch(url, {
method: "POST",
headers: {
"Authorization": `Bearer ${KEY}`,
"Content-Type": "application/json"
},
body: JSON.stringify({
messages: [{
role: "user",
content: "What's in our docs?"
}],
use_knowledge_base: true,
stream: true
})
});Your product.
Your knowledge. Your agent.
Atlas is live and ready. Let's plug it into whatever you're building.