Not just another layer of SaaS.Execution, Memory, Gateway, Billinguseful on their own, stronger together — compose intelligence like a multi-bit screwdriver.
curl https://api.lurus.cn/v1/chat/completions \
-H "Authorization: Bearer $LURUS_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"deepseek-chat","messages":[{"role":"user","content":"你好"}]}'One key for 50+ models, OpenAI SDK compatible.
Kova WAL-First engine with automatic crash recovery.
MemX adaptive memory with zero-LLM-cost distillation.
Lumen Replay + crash recovery + cost tracking.
Lucrum generates vnpy strategies from natural language and backtests them.
Switch unifies management of multiple AI CLIs, MCP, and cost.
Product MCP, Switch built-in servers, client directory.
Pinpoint common issues like 401 / quota / timeout on one page.
Picking the wrong model costs 10x more than writing the wrong code. Run one call through our gateway first, then decide whether to migrate.
The code is already written for you. Download and run — not a single line of config.
This isn’t a tool purchase — it’s an infrastructure replacement. Look at TCO and compliance boundaries first, then features.
The four foundational components of an LLM application: execution · memory · gateway · CLI. Each works on its own, stronger together.
Four decision points — not just another tool, but an infrastructure replacement.
From the Rust execution engine to the Flutter mobile client, all built in-house. Accounts / billing / memory / gateway are shared.
Kova scheduling at 3μs (Criterion benchmark) · 315K ops/s throughput.
Deploy once, pay no tax to any cloud vendor. End-to-end encryption with SM4-GCM.
Unified billing in 鹿贝 units — 50+ models, one statement.
| Dimension | Lurus Advantage | vs. Building It Yourself |
|---|---|---|
| Fully In-house Stack | From the Rust execution engine to the Flutter mobile client, core tech is fully self-controlled | Multi-vendor assembly, version lock-in risk |
| Engine-grade Performance | Kova 3μs scheduling latency, 315K ops/s (Criterion benchmark), zero external dependencies | Self-built Temporal/LangGraph are an order of magnitude slower |
| Data Sovereignty | On-premise deployment, data never leaves the enterprise boundary, supports SM4-GCM | Public cloud compliance and auditing are difficult |
| Unified Identity | All products share SSO, Passkey, MFA, with enterprise IdP federation | Self-operated Keycloak / Auth0 |
| Ecosystem Synergy | 12 products share accounts/billing/memory/LLM gateway — it compounds as you use it | Fragmented tool stack |
| Cost-efficient | MemX zero-LLM-cost distillation; Lucrum Decimal.js full precision with zero error | Memory / precision issues require extra investment |
| Transparent TCO | Unified billing in 鹿贝 units, pay-as-you-go + free quota | Reconciling bills across multiple vendors is complex |
| Open & Exportable | OpenAI SDK compatible, PAT/JWT standard authentication, export with no lock-in | High vendor lock-in exit cost |
Change one base_url and all your existing calls connect. One key for 50+ models, free quota on sign-up.