Workspace Understanding
@ references notes, sources, datasets, and files to focus agents contextually on what matters.
Upload, Edit, and Organize Files
Originals stay on your device. Agents process and index content locally with full audit trails.
Context Engine
Built-in guardrails and auto-checks keep agents grounded and evidence-driven with every decision.
Agent Orchestration
Routes work to the right agents, handles delegation, and surfaces the right model and tools for every task.
Multi-Hop Reasoning
Agents chain steps across multiple sources for complex tasks that need more than a single lookup.
Trusted Assistance
Focus on what only you can do. Delegate time-heavy tasks to agents that verify and cite information.
Atomic Evidence
Each note can contain atomic units, each with a unique highlight. Agents use these atomic highlights to form evidence-rich answers.
Automatic Bibliography
Sources are saved and traced automatically, building a ready-to-use bibliography with standard format options (APA, MLA, Chicago).
Traceable Generation
Ubik Agents generate cited output from your notes and documents with exact attribution so every claim links back to its source.
Agent + Human Annotation
Agents handle bulk annotation; you add judgment where it matters. A mix of both.
Click-Through Citations
Every claim links to a specific quote. Click any citation to jump straight to the source.
Search by Meaning
Find documents and passages by concept, not just keywords. Plain language works.
Deep Research
Agents fan out across academic databases, the web, and your own notes in one search. Results come back synthesized and cited, not as a raw list.
Source Scoring
AI evaluates and ranks sources by relevance, credibility, and citation impact to surface the best results.
Academic Databases
Query Google Scholar, Semantic Scholar, arXiv, PubMed, and more from a single search. No tab-switching, no lost results, no duplicate effort.
Full-Text Download
Ubik resolves open-access versions automatically. If full text is available through open access or your institution, it downloads directly to your workspace.
Personal Knowledge Base
Your own notes, highlights, and documents are searched alongside external sources. Prior work is always part of the picture.
Information Literacy
Papers download automatically and get indexed as annotatable, citable sources. Each search builds a knowledge base, not a link list.
Forced Interruption
At every consequential step, the agent stops cold. A card surfaces exactly what it plans to do, why, and with which parameters. Approve, edit, or reject before it moves.
Autonomy Levels
Dial in the right balance of oversight and automation. Choose from Full Spectrum, Writing Agent, Code Review, or Binary scales per workflow.
High-Stakes Only
Agents handle low-stakes steps automatically. Approvals are reserved for actions that change something: writing, querying external sources, or making irreversible calls.
Research Gate
Before hitting external databases, you see the full search scope: task, mode, and sources. Refine or redirect before the agent queries a single paper.
Document Brief
Before the agent writes or edits, you review the full brief: title, task, priority, and context. Change anything before it starts, not after.
Clarification Requests
When agents hit genuine uncertainty, they ask rather than guess. You provide the input; they proceed. Every exchange is logged and traceable.