Research and innovation
Find funding, build consortia and write stronger applications, with an agent working alongside your team.
ResearchArk- Funding discovery and intelligence
- Consortium building
- Application writing
- Research impact analytics
The root systemfor intelligentinstitutions.
Mycel designs, builds and runs AI platforms, models and agent systems for public institutions, research organisations and enterprises. Sovereign by design.
Platforms in production
ResearchArk, ArkNexus, URBNIA Indus and Mycel Noa.
Sovereign by design
EEA hosting and zero-retention inference for public-sector work, with human sign-off.
Funding intelligence
10,000+ funding organisations mapped in ResearchArk.
Four domains, one foundation
Every platform runs on the same foundation: our own agent runtime, our own models and one governance layer.
Public institutions
Run funding programmes, evaluations and AI governance with full transparency and a complete audit trail.
ArkNexus and URBNIA Indus- Call management and evaluation
- Expert and partner matching
- Public AI registers
- Ethics review and oversight
Enterprise
AI employees, private knowledge systems and transformation programmes that work inside your rules.
Mycel Noa, Mycel Meet and our services- Managed AI employees
- Private knowledge and search
- Sovereign meetings and translation
- Strategy, training and AI Act readiness
Models and agents
Our own models and the agent runtime every Mycel platform is built on.
Numo-1 and Ark-1- Scientific language models
- Agent-orchestration models
- High-performance agent runtime
- Open-source tooling
An AI-native operating model
Every person at Mycel works with a fleet of AI agents. Agents carry the volume, people make the decisions.
- BriefA request, a dataset or a question arrives.
- Agents workDrafting, checking, cross-referencing and monitoring run in parallel.
- People decideA specialist reviews, corrects and signs off.
- DeliveredShipped with a full audit trail.
SpecialistsData and AI engineering, research, scientific evaluation and go-to-market.
Engineering partnersDelivery capacity that scales with each engagement.
European infrastructureSupercomputing for training, EEA hosting for inference.
Benchmarks
How our runtime, models and data perform.
Mycel agent runtime
The orchestration core behind our agent systems, benchmarked against a widely used open-source agent SDK. Ranges cover every workload in the benchmark.
Faster on offline workloads3.2× to 7.8×
0×5×10×
Less memory46% to 64%
0%50%100%
Less CPU69% to 97%
0%50%100%
All injected faults handled. Benchmark report available on request.
Numo-1
Our open model for screening scientific literature, accurate on a PubMed benchmark and fast on a single consumer graphics card.
94.3%
Hugging Face, Apache-2.0, DOI 10.57967/hf/11188
ResearchArk
A funding graph covering 10,000+ funding organisations, with an agent that works across it.
10,000+
funding organisations. Get started
European hosting
EEA hosting for public-sector platforms.
Zero-retention inference
For public-sector work, model providers keep no data after answering.
Flexible deployment
Managed, sovereign cloud or on-premise.
Human oversight
Built for the EU AI Act and GDPR, with a person behind every external action.
From the blog
Research on who controls the AI stack, and on how work changes when AI does it.
India, Europe, and the Race to Build a Third Force in Global AI
India now has sovereign foundation models trained from scratch on 22 Indian languages, 38,000+ GPUs, and the fastest-growing AI hiring rate in the world. Europe has ASML, the AI Act, and €95.5B in research funding. Neither the US nor China has any incentive to connect them. That's exactly the opportunity.
The Full AI Stack — India, Europe, and the Race to Build a Third Force
The infrastructure for AI-driven work exists. The question is who controls the stack. This white paper maps the full seven-layer AI stack, assesses India's AI momentum and Europe's complementary strengths, and makes the case for a third force in global AI built not by outspending the superpowers, but by out-connecting them.
The Seven Layers of AI: Why Geopolitics Follows Architecture
Who controls AI isn't decided by which country has the best model. It's decided by who owns the energy, the chips, the data, the orchestration, and the governance layer above it all. A seven-layer map of the full AI stack — and where the real chokepoints sit.
Work with Mycel
Public institutions, research organisations, enterprises and investors.
Governments and public funders
Funding programmes, evaluations and AI governance, run inside the EEA.
Start a public-sector conversationResearch organisations and enterprises
Funding intelligence, custom agents and AI transformation, ready for the AI Act.
Book a working sessionInvestors and partners
Build with us on platforms and models.
Start a conversation