Inspectability
Systems that authorize people and AI should expose their boundaries, decisions, and evidence paths to review.
About
Technology scales when the system around it works
Years across complex enterprise projects, digital transformation, and demanding production—including automotive—showed that adoption needs more than technical capability: stability, clear processes, accountable resources, economic visibility, and auditability. CanDoItAll connects AI and agents to that operating context instead of leaving them as another isolated tool.
Why open source
Useful know-how should compound
Organizations repeatedly solve the same orchestration, integration, and governance problems behind closed doors. Publishing the core makes the architecture, trade-offs, and reusable patterns inspectable.
Sharing the implementation allows practitioners to challenge assumptions, improve the foundations, and reuse focused modules without surrendering control of their deployment or data.
Shared know-how
Reusable engineering patterns should save the next team from rebuilding the same foundations in isolation.
Deployment sovereignty
A local-first, self-hostable core lets organizations retain control of infrastructure, credentials, and operational data.
Technical foundation
From industrial automation to governed AI
CanDoItAll is shaped by systems that had to coordinate machines, software, data, and accountable decisions—not by an isolated AI demo.
View LinkedIn profile-
X-ray, CT, and visual inspection
Before CanDoItAll, Tomas built and ran a company developing its own industrial inspection systems for manufacturers, including automotive environments.
-
Production-line control
That work included modular control software for robots, lasers, pallet systems, inspection equipment, and operator workflows—supporting architectures from a single standalone station to distributed production lines.
-
Digital transformation
Several years of work with larger companies added experience with modernization, integration boundaries, organizational adoption, and long-lived enterprise systems.
-
Applied AI over time
Early experiments with conversational systems led into bioelectronics and bioinformatics studies with neural networks, then industrial machine learning for imaging, vibrodiagnostics, and product recognition. LLMs opened the next chapter: agents, tool use, and governed automation.
-
A project operating system
CanDoItAll brings those lessons together: modular contracts, explicit responsibility, controlled automation, traceability, and evidence around the work being delivered.
Sustainable by delivery
Keep the core open — fund the work through real adoption
Continued development is supported through business cooperation around successful implementation—not by closing the reusable foundation.
Organizations can fund work they need while improvements that belong in the shared core remain available to the wider project.
Explore business cooperationAdoption and architecture
Map CanDoItAll into existing governance, data, security, and operating models.
Integration and tailored capability
Connect enterprise systems or build focused modules without weakening project boundaries.
SLA and operational support
Define response, maintenance, and continuity arrangements for serious deployments.
Knowledge transfer
Help internal teams understand, operate, extend, and govern the system themselves.
Project timeline
From first commit to a stable 1.0
The roadmap records why the project began, what has opened, and what needs to become reliable next. Future dates are current targets and may move as the alpha produces better evidence.
-
March 2026 Project started
I had wanted a system like this for years. Managing many simultaneous projects across separate tools made context difficult to keep clear and maintain. Direct integration of AI and agents suggested a better model: one operating system around project intent and execution. AI coding tools made the scale feasible, but this was clearly a long-term project from the start. I chose open source because shared knowledge reduces duplicated complexity and helps the wider ecosystem advance.
-
August 2026 Public repositories opened
After five months of development and experimentation around AI, agents, and the shape of the platform, I opened the repositories. It is an early milestone: parts of the code still need refactoring, hardening, and clearer setup. Even so, the foundation is becoming stable enough for technical exploration and outside feedback.
-
October 2026 Main refactor and stabilization target
The next phase concentrates on modules that are still sketched rather than fully productized. Refactoring and hardening those foundations comes before adding more breadth. Once that work is complete, the platform can add more ready-to-use AI providers, memory integrations, and plugins on top of clearer contracts.
-
November 2026 Economy module target
CanDoItAll already records agent prices and worker costs from CRM/HR, and uses previous runs to estimate workflow and process costs. Today, most of that visibility focuses on AI execution and labour. Deeper integration with Ledger and the economy module is intended to expand project-level analysis, planning, simulation, and execution.
-
December 2026 Version 1.0 target
The main target is a stable 1.0 of the core application by year-end: easier setup, hardened modules, and as many high-value defects resolved as possible. An early alpha or beta of Cognitive Memory may accompany it, but the reliability of the core application remains the release priority.
-
2027 Roadmap shaped by real use
Feedback from alpha and beta use will shape the 2027 roadmap more than assumptions can. A known focus is deeper Cognitive Memory; several additional modules remain intentionally unannounced until their boundaries and value are clearer.
Contact