Universal SLR Assistant
sadeghanisi.github.io
- Free plan
- Open source
- Windows
- macOS
- Linux
- Web
- API
- Self-hosted
What Universal SLR Assistant is
Universal SLR Assistant supports systematic and scoping reviews with AI-assisted screening, extraction, and reproducible outputs. It offers a desktop GUI and a browser-based local Web App. References can be imported from RIS, BibTeX, CSV, and plain-text files, including exports from PubMed, Scopus, and Web of Science. Its workflow spans abstract screening, full-text PDF processing, structured extraction, and export; likely duplicates are identified through DOI and fuzzy title matching. Workspace mode stores references, PDF metadata, review items, decisions, and audit events in a local SQLite workspace, while keeping AI suggestions distinct from human decisions. Workspace exports can include decision CSV/XLSX files and logs, exclusion reasons, PRISMA-ready counts, a methods disclosure draft, and a manifest. Provider options include cloud services and local endpoints such as Ollama, as well as custom OpenAI-compatible endpoints. Cloud providers receive submitted paper text; local endpoints can keep it on the user’s computer or network. AI output is assistance only: people remain responsible for verification and final review decisions. The free, MIT-licensed plan is free; cloud AI provider API fees may apply.
Who it suits
It is intended for PhD students, researchers, professors, evidence synthesis teams, and authors conducting systematic or scoping reviews. It may suit reviewers who need structured exports and want to distinguish AI suggestions from human decisions.
Good to know, and what to watch
- Imports RIS, BibTeX, CSV, and plain-text references.
- DOI and fuzzy-title matching identify likely duplicates.
- Workspace stores review data in local SQLite.
- Exports include PRISMA-ready counts and decision logs.
- Supports local and cloud AI provider options.
- Cloud providers receive submitted paper text.
- Cloud AI provider API fees may apply.
- Human verification and final review decisions remain required.
- Local Web App is not intended as a public multi-user service.
Plans
Every plan and priceFree and open source
Free plan
Getting started
- Open the project website
- Choose the desktop GUI or local browser-based Web App
- Import RIS, BibTeX, CSV, or plain-text references
- Select a cloud, local, or custom OpenAI-compatible AI provider
- Process abstracts and full-text PDFs, then perform structured extraction
- Export decisions, logs, exclusion reasons, counts, methods, or a manifest
Limits to check
The software is free, but cloud AI provider API fees may apply. AI results are assistance only, so people must verify outputs and make final inclusion, exclusion, extraction, and reporting decisions. The local Web App binds to 127.0.0.1 and is not intended for public multi-user use.
The Club's review
Universal SLR Assistant is a free, open-source workspace for systematic and scoping reviews, covering screening, full-text work, extraction, and reporting. It best suits researchers and evidence-synthesis teams that want local records and flexible AI-provider choices. The trade-off is clear: people remain responsible for checking outputs and making every substantive review decision.
Overview
The workflow starts with reference import and continues through abstract screening, full-text PDF processing, structured extraction, and export. RIS, BibTeX, CSV, and plain-text files are accepted, including exports from PubMed, Scopus, and Web of Science. DOI matching and fuzzy title matching identify likely duplicates, reducing avoidable screening work while leaving the final judgment to the reviewer.
Workspace mode stores references, PDF metadata, review items, decisions, and audit events in a local SQLite database. AI suggestions are kept separate from human decisions, an important distinction when a review needs an accountable record rather than an opaque model verdict. Exports cover decision CSV and XLSX files, AI and human decision logs, full-text exclusion reasons, PRISMA-ready counts, a methods disclosure draft, and a manifest.
Key features
Broad AI choice
Provider options include OpenAI, Anthropic Claude, Google Gemini, Ollama, OpenRouter, DeepSeek, Mistral, Kimi, xAI, LM Studio, vLLM, LocalAI, and custom OpenAI-compatible endpoints. That range lets a team align the workflow with an existing account, a self-hosted service, or a local model. It also means provider setup and model governance become the user's responsibility rather than a single vendor's managed experience.
Local workspace and privacy controls
The desktop GUI and browser-based local Web App are designed for work on the user's machine. The Web App binds to 127.0.0.1 and is not intended to be a public, multi-user service, so it is a poor choice for a shared online collaboration hub. API keys are not written to desktop or Web App settings files, and audit ledgers exclude keys and full paper text. Cloud providers still receive submitted paper text; Ollama or another local endpoint can keep that text on the computer or local network. Choose the provider model with the sensitivity of the corpus in mind.
Human-controlled review stages
Screening, paper summaries, full-text analysis, data extraction, and storage are all covered, but AI output is assistance only. Human verification and final inclusion, exclusion, extraction, and reporting decisions remain required. This is appropriate for defensible evidence synthesis, not for automating a review from submission to publication.
Pricing
The Free and open source plan costs nothing under the MIT License. There is no paid tier or seat schedule stated. Cloud AI provider API fees may apply, so the software itself can be free while model usage creates a separate bill. The plan is a strong fit for students, independent researchers, and teams able to manage their own provider accounts; organizations wanting bundled usage, vendor support, or centralized administration should look elsewhere.
Platforms
Universal SLR Assistant supports Linux, macOS, Windows, and the web through its local browser-based app. The combination is useful for researchers who need a desktop interface or a local browser workflow, but the Web App's loopback-only design limits it to local use rather than public deployment.
Who it's for
PhD students, researchers, professors, evidence-synthesis teams, and systematic or scoping review authors are the natural audience. It is especially well matched to projects requiring an audit trail, reproducible exports, and control over whether paper text goes to a cloud or local endpoint. It is less suitable for casual literature browsing, teams seeking a hosted multi-user service, or anyone unwilling to verify model-assisted decisions.
Pros and cons
Pros
- End-to-end review coverage: import, screening, PDF processing, extraction, and export sit in one workflow.
- Traceable outputs: separate AI and human logs, exclusion reasons, PRISMA-ready counts, a methods draft, and a manifest support reproducibility.
- Provider flexibility: cloud, local, and custom OpenAI-compatible options accommodate different budgets and data policies.
- Local control: SQLite storage and loopback Web App operation keep the workspace on the user's machine.
Cons
- Human work remains essential: the tool cannot be treated as an authority for final review decisions.
- Cloud privacy depends on provider choice: submitted paper text goes to cloud services unless a local endpoint is used.
- Not a public collaboration service: the local Web App is not intended for multi-user deployment.
- Usage costs can sit outside the software: cloud provider API fees may apply despite the free licence.
Alternatives
For a broad category view, see Systematic Review Software and AI Literature Review Software.
- RevMan is the better comparison for users who want a freemium product with a free trial across desktop and web platforms.
- ReviewAid suits readers seeking a free, open-source option with self-hosted and web availability.
- Rayyan is preferable when mobile apps, an API, and a hosted web service matter, though its free plan limits active reviews and reviewers.
- Silvi is worth choosing for a web-first workflow with unlimited reviews on its Basic plan and paid rapid features.
- Covexe fits teams wanting a hosted freemium service with an AI co-pilot allowance.
- Parsifal is a free web option where AI is optional and uses the user's Anthropic or OpenAI key.
- Systematicly is aimed at users who want free personal drafts and limited Evidence Answers before upgrading.
- Nested Knowledge is the paid alternative for teams considering academic per-user plans and a free trial.
Verdict
Choose Universal SLR Assistant if you need a no-cost, open-source workflow that preserves human control, records review activity, and offers both cloud and local AI routes. Look elsewhere if you need hosted multi-user collaboration, bundled AI usage, or a system that makes final decisions without ongoing expert verification.
The verdict
Universal SLR Assistant covers the main review stages and provides structured workspace exports. Its AI outputs are not final decisions, and users should choose a provider with the paper-text transmission model in mind.
Questions people ask
Is Universal SLR Assistant free?
Yes. The Free and open source plan costs nothing and uses the MIT License. Cloud AI provider API fees may apply.
Which review types does it support?
It supports systematic reviews and scoping reviews.
Which AI providers can it use?
Options include OpenAI, Anthropic Claude, Google Gemini, Ollama, OpenRouter, DeepSeek, Mistral, Kimi, xAI, LM Studio, vLLM, LocalAI, and custom OpenAI-compatible endpoints.
Can it run locally?
Yes. It offers a desktop GUI and a browser-based local Web App. Ollama or another local endpoint can keep paper text on the user’s computer or local network.
What files can it import and export?
It imports RIS, BibTeX, CSV, and plain text, including PubMed, Scopus, and Web of Science exports. Export formats include Excel, CSV, JSON, Markdown, and plain text.
Are AI decisions final?
No. AI output is assistance only; human verification and final inclusion, exclusion, extraction, and reporting decisions are required.
What the maker publishes
- Abstract screening
- AI evaluates titles and abstracts against user criteria and returns Include, Exclude or Maybe recommendations for human review. source
- AI providers
- Provider options include OpenAI, Anthropic Claude, Google Gemini, Ollama, OpenRouter, DeepSeek, Mistral, Kimi, xAI, LM Studio, vLLM, LocalAI, and custom OpenAI-compatible endpoints. source
- API and extensibility
- The Web App exposes workspace review and export HTTP endpoints, and contributors can add providers through the LLM interface or OpenAI-compatible profiles. source
- Audit ledger
- Each run can create a JSONL ledger containing provider, model, hashes, cache status, parsing status and retry metadata without storing API keys or full paper text. source
- Caching
- Configuration-aware JSON caching keys results by text, stage, provider, model, prompts and extraction settings. source
- Cloud data transmission
- When cloud providers are selected, submitted paper text is sent to the provider's API; router providers add another intermediary. source
- Data transmission
- Cloud providers receive submitted paper text, while Ollama or another local endpoint can keep paper text on the user's computer or local network. source
- Deduplication
- Likely duplicate references are identified using DOI matching and fuzzy title matching. source
- Deduplication
- Duplicate detection combines DOI matching with fuzzy title matching and leaves final review to the researcher. source
- Developer post: automation
- A mirrored developer post lists automated deduplication and extraction as product capabilities. source
- Developer post: cost
- A mirrored developer post says the tool has no monthly fees and is free under the MIT license. source
- Developer post: local integration
- A mirrored developer post highlights local Ollama integration and processing PDFs on the user's own machine. source
- Developer post: platforms
- A mirrored developer post names Windows, macOS and Linux as supported desktop platforms. source
- Developer post: privacy
- A mirrored developer post claims local LLM use can keep data from leaving the user's hardware. source
- Developer post: providers
- A mirrored developer post describes support for more than nine AI providers, including OpenAI, Claude, Gemini and DeepSeek. source
- Developer profile
- A public profile identifies Mo Anisi as holding a BSc in Biology, an MSc in Cognitive Science and an MSc in Health and Social Welfare. source
- Documentation
- Documentation includes a 1,550-line complete user guide, in-app help, troubleshooting guidance, tests and GitHub issues. source
- Exports
- Workspace exports include decision CSV/XLSX files, AI and human decision logs, full-text exclusion reasons, PRISMA-ready counts, a methods disclosure draft, and a manifest. source
- Exports
- Workspace exports include decision CSV/XLSX files, AI and human decision logs, exclusion reasons, PRISMA-ready counts, a methods draft and a manifest. source
- Hardware guidance
- The user guide lists 4 GB RAM and 2 GB disk as minimums, with 8 GB RAM and 5 GB disk recommended; Ollama guidance lists at least 8 GB RAM and 10 GB disk. source
- Human decisions
- AI outputs remain suggestions, while final Include, Exclude or Maybe decisions require a human reviewer. source
- Installation
- Installation uses a Git clone, virtual environment and pip requirements, with batch launchers for Windows and Python commands for macOS or Linux. source
- Intended users
- It is intended for PhD students, researchers, professors, evidence synthesis teams, and systematic or scoping review authors. source
- Interfaces
- The project provides a desktop GUI and a browser-based local Web App. source
Checked against the maker's pages on 15 Oct 2026.