Local-first research memory and evidence infrastructure for researchers and AI agents.
Research evolves across conversations, repositories, notebooks, papers, and experiments. RKA helps that work survive across tools, agents, and sessions by preserving not only findings, but also the decisions, evidence, provenance, and reasoning needed to reconstruct a complete research story.
- Capture observations, literature, decisions, experiments, failures, and open questions as the work happens.
- Crystallize noisy records into linked claims, evidence, research questions, and traceable conclusions.
- Reuse the resulting knowledge to resume work, recover the complete reasoning behind an idea, and support future research outputs.
The canonical research knowledge layer. RKA Core provides local-first storage, structured retrieval, provenance, integrity checks, backup and recovery, and project continuity through MCP, REST, CLI, and a local web dashboard.
Status: Available and under active development. Core reliability and install-friendly distribution are the current priorities.
A separate researcher-in-the-loop workbench built over RKA Core. Writer is designed to help researchers shape an insight into a coherent argument, discuss and revise the paper spine, and draft from selected evidence while keeping framing, terminology, emphasis, and final prose under researcher control.
Status: In development.
- Install and run RKA Core
- Read the Core documentation
- Explore RKA Writer
- View the ecosystem roadmap
- Visit the project website
- Local-first: research records remain under the researcher’s control.
- Traceable: important claims and decisions retain their evidence and provenance.
- Researcher-controlled: AI suggestions do not silently become canonical research knowledge.
- Recoverable: durable state can be inspected, exported, backed up, and restored.
- Provider-agnostic: agents and applications integrate through explicit, model-independent interfaces.
- Modular: knowledge infrastructure and downstream research tools evolve as separate products.
We welcome researchers, developers, and tool builders interested in durable AI-assisted research workflows.
- Report a Core issue
- Read the Core contributor guide
- Discuss Writer requirements
- Review the public roadmap
RKA Project is developed by researchers for research workflows where continuity, provenance, and human judgment matter.