About

ALYSIS AI I.K.E. is an Athens-based deep-tech company building AI tools that make generative AI practical, affordable and genuinely usable. Our name comes from the Ancient Greek word for chain: the linked steps that carry a request from plan to shipped, reviewed code. Our first product is Alysis Code (formerly Sylliptor), an open-source, end-to-end AI coding agent released under Apache 2.0 and publicly available today. It runs in a developer's terminal and takes on the work that takes hours rather than minutes: full project scaffolding, large multi-file refactors, and complete features built from a description. Around it we are building a wider suite. More open-source developer tools are in progress, alongside Gen AI Services: custom and open-source model APIs, model hosting, and fine-tuning, delivered as managed services for teams that want production AI without building the operations around it. The problem we address is cost and access. Serious use of generative AI is expensive and operationally complex, which locks out exactly the smaller teams that would benefit most. We keep our tools open source, free of vendor lock-in, and independent of any single model provider, so customers can inspect what they run, choose where it runs, and stay in control of what they pay.

Solution & Product

Alysis Code is our live product: an open-source AI coding agent that runs in the developer's terminal and handles work measured in hours rather than minutes, such as full project scaffolding, large refactors across many files, and complete features built from a description. Its core is Forge, a plan-driven execution mode. Given a request, Forge asks clarifying questions when the ask is vague, writes an explicit plan of scoped tasks with runnable file boundaries, dispatches parallel workers to execute them, verifies each task before marking it done, and records every failure as a structured issue the next run avoids. Plans, traces, and per-task artifacts persist on disk, so any run can be resumed. It is provider-independent by design. Teams bring their own model (OpenAI, Anthropic, DeepSeek, Qwen, Gemini, Mistral, xAI, Moonshot, OpenRouter and others) or point it at a local endpoint such as Ollama, LM Studio or vLLM and run fully offline. Different roles can use different models: planner, coding workers, reviewer. Safety is default-on: sandboxed execution via Docker or Bubblewrap, an always-on command denylist, approval gates for shell and Git, and complete session logs for review and audit. It ships from PyPI via a single pipx command, with a VS Code extension, self-hosting, and an enterprise path offering private deployment, SSO, and org-wide policy. Licensed Apache 2.0.

Innovation & Scalability

Three things distinguish our approach. Plan-driven, host-verified execution. Most coding agents optimise a single turn. Forge separates planning from implementation, runs implementation across isolated parallel workers, and treats host-side evidence, meaning real diffs and real test runs, as authoritative rather than trusting the model's own account of what it did. Cross-run memory converts each failure into a constraint the next attempt respects, so the agent improves within a project instead of restarting cold every time. Model independence. We are not bound to one provider's pricing or roadmap. Per-role routing lets the planner, the coding workers and the reviewer each use a different model, and support for local endpoints means the same workflow runs on a frontier API, on cheap open-weights models, or entirely on-premise for regulated work: a hedge against both cost shocks and data-residency constraints. An extensible platform, not a closed product. Tools, skills, hooks, MCP servers and custom subagents let teams shape the agent around their own stack, with sandboxing and approval gates on by default. Scaling is software-led. The tools are open source and self-serve, so distribution costs almost nothing; the managed services are API-delivered; and each new customer adds load we meet with capacity rather than headcount.

Traction & achievements

Product. Alysis Code is publicly available and open source under Apache 2.0, out of closed beta. The CLI ships from PyPI at version 0.13.0 with a companion VS Code extension. It supports twelve model providers plus local endpoints, and includes Forge plan-driven execution, parallel workers, sandboxing, approval gates, cross-run memory, and an extensible layer of tools, skills, hooks, MCP servers and custom subagents. Commercial foundation. Billing is implemented end to end on Stripe with a credit-based model: free monthly credits, a bring-your-own-key route at no cost, and three paid tiers. Account management, device-code CLI sign-in, and an enterprise offering covering private deployment, SSO, role-based access and a session audit trail are built and in place. Engineering. Beyond the agent we have built a model training and fine-tuning platform that trains and optimises models reliably at scale. It underpins the Gen AI Services now in development. Company. ALYSIS AI I.K.E. is incorporated in Athens with a four-person core team, funded to date by the founders and angel investment. We operate two public properties: alysisai.com for the company, and a dedicated product site with full documentation for Alysis Code.

Team & Why Us

Team. Three co-founding engineers, plus research and marketing leads. Apollon Foivos Bakis (Co-Founder, Lead Engineer): MSc Quantitative Finance, University of Manchester; former finance domain expert and peer reviewer at Mercor, training AI models for frontier labs. Andreas Klimentidis (Co-Founder, Senior Engineer): BSc Computer Science; years of production full-stack delivery. Panagiotis Perdikis (Co-Founder, Senior Engineer): final-year MEng in Mechanical Engineering at NTUA, focused on applied AI systems. Yulu Ding (Lead Marketing Strategist): MA University of Manchester, MSc in Data Science and Machine Learning in progress. Why us. We ship. The agent is built, released and running; billing, authentication and the enterprise path are in place. Everything user-facing is Apache 2.0, so a prospective customer can read the code before deciding rather than take our word for it. Growth strategy. Open source is the top of the funnel: developers install free and convert to hosted credits when they want capacity without managing their own API keys. Those developers carry the tool into their companies, where paid seats and enterprise deployments follow. Gen AI Services then extend the same relationship from tooling into managed model APIs, hosting and fine-tuning.

Funding Needs

We are raising a seed round to take Alysis Code from a released product to a commercial one. Use of funds. The largest share goes to engineering: growing the core team beyond its current four so we can ship the wider tool suite and move Gen AI Services, meaning custom and open-source model APIs, model hosting and fine-tuning, from development into general availability. A significant share goes to compute and infrastructure: GPU capacity, hosting, and the training and evaluation cycles those services depend on. The remainder funds go-to-market, meaning developer relations, documentation, content and community work, which is how open-source tools actually reach their users, together with the legal, security and compliance work an enterprise offering requires. What it buys. Runway to convert open-source adoption into paying accounts, land the first enterprise deployments, and bring the managed services to revenue. We are equally open to non-dilutive support: Greek and EU research and innovation funding, and partnerships with cloud or model providers whose reach our tooling extends. Beyond capital, we are looking for investors and advisors with a track record in developer tools, open-source commercialisation, and enterprise software sales in Europe.