Eldric is the distributed agentic AI platform for serious deployments. Built in Austria — the birthplace of LSTM and xLSTM (Sepp Hochreiter). From a single laptop to a 50-node cluster, from local Ollama to cloud APIs — one platform, fully under your control.
Multi-tier architecture with specialized workers, intelligent routing, and horizontal scaling.
Multiple LLMs reasoning together as a collective intelligence. Auto-selects the best strategy based on your query.
Models argue positions across rounds. A judge renders the final verdict. Best for decisions.
Model A generates, Model B critiques, A refines. Iterates for quality. Best for writing.
Fan-out to N models, judge picks the single best answer. Best for code generation.
All models answer, consensus analysis with confidence. Best for factual questions.
Merge insights from all models into one comprehensive answer. Default strategy.
Auto-Selection
The router automatically picks the optimal strategy based on query content: “Kafka vs RabbitMQ?” triggers debate, “Write a migration plan” triggers critique, “Is this thread-safe?” triggers vote. Override with ensemble_strategy in the API.
Each agent constrains available tools for focused, safe execution.
Everything you need to deploy, scale, and operate AI infrastructure.
Ollama, vLLM, TGI, llama.cpp, MLX, OpenAI, Anthropic, Azure, Groq, xAI, Together, Mistral, Cohere, NVIDIA Triton, TensorFlow Serving, TorchServe, ONNX Runtime, and more. Switch backends without changing code.
FreeIntelligent routing based on model specialization, load metrics, latency, and query intent. The router uses an LLM for real-time decisions, with optional xLSTM predictor for workload forecasting, anomaly detection, and fast classification.
EnterpriseBuilt-in vector database with SQLite, FAISS, or ChromaDB backends. Automatic text chunking, embedding generation (Ollama, OpenAI), and semantic search. Multi-tenant with namespace isolation.
StandardTLS termination, API key auth, rate limiting, embedded web chat UI with PWA support. Proxies worker registration for geo-distributed clusters. Built-in NAT tunnel for workers behind firewalls — no VPN needed.
FreeSpread workers across university labs, corporate datacenters, home offices, and cloud providers. Workers register through the Edge TLS gateway over the internet. Mix RTX 3090s, H100s, and Apple Silicon in the same cluster. Architecture →
StandardWorkers behind NAT/firewalls receive inference requests through outbound-only long-poll connections to the Edge. No inbound ports, no VPN, no public IP. Just --tunnel https://edge.example.com. Edge docs →
Autonomous goal decomposition, multi-agent collaboration, agent marketplace, human-in-the-loop approval, and visual swarm dashboard with real-time topology.
ProfessionalLoRA, QLoRA, SFT, DPO, RLHF via Unsloth, Axolotl, TRL, DeepSpeed, MLX. Native xLSTM training backend for Hochreiter's extended LSTM architecture. Training chains with visual node editor. Latent reasoning: COCONUT, Quiet-STaR, Pause Tokens, Hidden CoT, DeepSeek DSA.
StandardSpeech-to-text (Whisper.cpp, OpenAI), text-to-speech (Piper, ElevenLabs), voice chat, video transcription, scene detection, multimedia RAG indexing.
Free (basic)Email (IMAP/SMTP), SMS (Twilio), WhatsApp (Business API), Signal (E2E encrypted), Microsoft Teams, XMPP. AI auto-response with approval workflow.
ProfessionalNASA, ESA, CERN, LIGO, JWST, PubMed, ClinicalTrials.gov, NCBI, UniProt, Materials Project, GBIF. Bioinformatics, CRISPR, pharmaceutical, LIMS with GLP/FDA compliance. xLSTM-powered anomaly detection on scientific datasets.
Free (basic)Extensible plugin architecture: tools, filters, widgets, integrations. Public catalog at eldric.ai/plugins with one-click install. Per-user preferences synced to server.
FreeRemote command execution, script deployment, log collection, rolling restarts, and system status across all cluster nodes via the Controller API.
EnterpriseSplit models larger than any single GPU across multiple workers. GGUF layer-level sharding with VRAM-aware scheduling. Automated llama.cpp RPC orchestration — deploy a 70B model across 3 workers with one API call. Learn more →
StandardDirect GGUF model loading via embedded llama.cpp — no Ollama, no vLLM, no external dependencies. Permanent VRAM residence for zero cold-start latency. CUDA + Metal GPU support, continuous batching, xLSTM models, pipeline parallelism, and Data Worker model distribution. Learn more →
FreeEd25519-signed licenses with hardware binding. Online/offline validation. Per-backend worker limits. Five tiers: Free, Standard, Professional, Enterprise, Custom.
Free tier availableAutomatic detection: native tool calling for supported models (Ollama 0.3+, OpenAI, Anthropic), XML fallback for others. 40+ built-in tools across all categories.
FreeC++ CLI for terminal, native macOS SwiftUI GUI, iOS app, embedded web chat, OpenWebUI compatible. All clients share the same distributed backend.
FreeAgent Builder: LLM-powered creative agent design. Agent Generator: template-based mass production. Both deploy directly to your cluster.
StandardNative support for Sepp Hochreiter's xLSTM architecture. Dedicated xLSTM training backend, xLSTM predictor in the router for workload forecasting and fast sequence classification, and xLSTM-based anomaly detection across workers.
StandardSmart home integration, Industrial IoT with OPC-UA protocol, real-time sensor monitoring, AI-powered anomaly detection on production data.
ProfessionalChoose the tier that fits your deployment. Free tier includes everything you need to get started.
| Feature | Free | Standard | Professional | Enterprise |
|---|---|---|---|---|
| Infrastructure | ||||
| Controllers | 1 | 1 | 2 | 5+ |
| Routers | 1 | 2 | 4 | 10+ |
| Inference Workers (Proxy) | 2 | 3 | 10 | 50+ |
| Native Inference Workers | 1 | 3 | 10 | Unlimited |
| Edge Servers | 1 | 2 | 5 | Unlimited |
| AI & Inference | ||||
| LLM Backends | All 20+ | All | All | All |
| Streaming SSE | ✓ | ✓ | ✓ | ✓ |
| Native Tool Calling | ✓ | ✓ | ✓ | ✓ |
| Swarm LLM (ensemble) | ✓ | ✓ | ✓ | ✓ |
| AI-Powered Routing | — | — | — | ✓ |
| Data & Knowledge | ||||
| RAG / Vector Storage | — | ✓ | ✓ | ✓ |
| Embeddings | — | ✓ | ✓ | ✓ |
| Database Connectivity | SQLite | +PG, MySQL | +DB2 | All |
| NFS Storage | — | — | ✓ | ✓ |
| Agents & Orchestration | ||||
| 13 Agent Types | ✓ | ✓ | ✓ | ✓ |
| Multi-Agent (parallel) | — | ✓ | ✓ | ✓ |
| Swarm Orchestration | — | — | ✓ | ✓ |
| Agent Builder | 3/month | 20/month | 100/month | Unlimited |
| Media & Communication | ||||
| STT (Speech-to-Text) | Basic | +Streaming | +Diarization | All |
| TTS (Text-to-Speech) | Basic | +Streaming | +Voice Clone | All |
| Comm Protocols | +SMS | All except Signal | All (incl. E2E) | |
| Science & Training | ||||
| Science APIs (140+) | Basic | +BLAST | +Docking, ADMET | All + AlphaFold |
| Training Backends | MLX only | +Unsloth, TRL, xLSTM | All | All + Distributed |
| xLSTM Predictor (Router) | — | ✓ | ✓ | ✓ |
| Training Chains | — | ✓ | ✓ | ✓ |
| LIMS / 21 CFR Part 11 | — | — | GLP | Full FDA |
| Operations & Security | ||||
| Plugin System | ✓ | ✓ | ✓ | ✓ |
| Cluster Ops (remote exec) | — | — | ✓ | ✓ |
| Dashboard & Metrics | — | — | ✓ | ✓ |
| Multi-Tenant Isolation | — | ✓ | ✓ | ✓ |
| Hardware Binding | — | — | — | ✓ |
| Priority Support | — | — | — | ✓ |
Connect to any LLM runtime — local or cloud. All backends support streaming.
| Backend | Streaming | Vision | Native Tools | Embeddings | Type |
|---|---|---|---|---|---|
| Ollama | ✓ | ✓ | ✓ | ✓ | Local |
| vLLM | ✓ | ✓ | ✓ | ✓ | Local |
| llama.cpp | ✓ | ✓ | — | ✓ | Local |
| MLX | ✓ | — | — | — | Local (Apple) |
| TGI | ✓ | ✓ | — | — | Local |
| NVIDIA Triton | ✓ | ✓ | — | ✓ | Inference |
| OpenAI | ✓ | ✓ | ✓ | ✓ | Cloud |
| Anthropic | ✓ | ✓ | ✓ | — | Cloud |
| xAI (Grok) | ✓ | ✓ | ✓ | — | Cloud |
| Groq | ✓ | — | ✓ | — | Cloud |
| Together AI | ✓ | ✓ | ✓ | ✓ | Cloud |
| HuggingFace | ✓ | — | ✓ | ✓ | Cloud |
Download Eldric and have your AI infrastructure running in minutes.
Download Eldric View Licensing