Search, cite, discover
- PubMed search — full-text search, citation graphs, MeSH lookup
- bioRxiv / medRxiv — preprint surveillance and alerts
- Scientific writing — claim-evidence aligned drafting
A paired scientist for your lab
From research question to evidence-backed work — without the busywork
sciClaw is not a "write my paper" bot. It's a paired-scientist agent: it proposes a plan, runs the right tools (literature, documents, code), cites sources, and logs decisions so you can reproduce and audit the work.
You stay in control: approve key steps, review outputs, and keep the evidence in your workspace.
Free and open source. MIT licensed. Works with your existing tools.
Sound familiar?
Research involves a lot of repetitive work that pulls you away from thinking.
Searching takes forever
Hours lost sifting through PubMed results, chasing citation chains, and formatting references by hand.
Too many tools, too much pasting
Copy data into Excel, export to Word, fix formatting, remake the figure… rinse and repeat for every revision.
No trail of what changed
Six months later, you can't reconstruct which analysis produced which result or why a parameter was chosen.
sciClaw handles the tedious parts so you can focus on the science.
See it in action
You type a question. sciClaw does the work.
$ sciclaw agent -m "What pathways are implicated in ALS progression?"
Literature search results (PubMed, 2020–2025):
1. TDP-43 proteinopathy — Cytoplasmic mislocalization of TDP-43 in >97% of ALS cases. RNA processing disruption leads to downstream neurodegeneration. (Neumann et al., 2023; PMID 37284100)
2. SOD1 / oxidative stress — Gain-of-function SOD1 mutations cause mitochondrial dysfunction and ROS accumulation in motor neurons. (Rosen et al., 2022; PMID 36198432)
3. C9orf72 repeat expansion — Hexanucleotide repeat in C9orf72 accounts for ~40% of familial ALS. Dipeptide repeat proteins impair nucleocytoplasmic transport. (DeJesus-Hernandez et al., 2021; PMID 34892710)
Found 47 articles. Full citation list saved to session.
1. TDP-43 proteinopathy — Cytoplasmic mislocalization of TDP-43 in >97% of ALS cases. RNA processing disruption leads to downstream neurodegeneration. (Neumann et al., 2023; PMID 37284100)
2. SOD1 / oxidative stress — Gain-of-function SOD1 mutations cause mitochondrial dysfunction and ROS accumulation in motor neurons. (Rosen et al., 2022; PMID 36198432)
3. C9orf72 repeat expansion — Hexanucleotide repeat in C9orf72 accounts for ~40% of familial ALS. Dipeptide repeat proteins impair nucleocytoplasmic transport. (DeJesus-Hernandez et al., 2021; PMID 34892710)
Found 47 articles. Full citation list saved to session.
$ sciclaw agent -m "Draft a methods section for our ERP correlation analysis"
Methods
Participants. Thirty-two right-handed adults (18 female; mean age = 24.3, SD = 3.1) participated after providing written informed consent. The study was approved by the institutional review board (Protocol #2024-0187).
ERP Recording. Continuous EEG was recorded from 64 Ag/AgCl electrodes using a BioSemi ActiveTwo system (sampling rate: 512 Hz). Data were referenced to the average of bilateral mastoids and filtered offline (0.1–30 Hz, zero-phase Butterworth).
Statistical Analysis. Spearman rank correlations were computed between P300 amplitude at Pz and ALSFRS-R decline rate (points/month). Multiple comparisons were corrected using the Benjamini-Hochberg procedure (FDR = 0.05).
Draft saved. Run "sciclaw agent" to continue refining.
Participants. Thirty-two right-handed adults (18 female; mean age = 24.3, SD = 3.1) participated after providing written informed consent. The study was approved by the institutional review board (Protocol #2024-0187).
ERP Recording. Continuous EEG was recorded from 64 Ag/AgCl electrodes using a BioSemi ActiveTwo system (sampling rate: 512 Hz). Data were referenced to the average of bilateral mastoids and filtered offline (0.1–30 Hz, zero-phase Butterworth).
Statistical Analysis. Spearman rank correlations were computed between P300 amplitude at Pz and ALSFRS-R decline rate (points/month). Multiple comparisons were corrected using the Benjamini-Hochberg procedure (FDR = 0.05).
Draft saved. Run "sciclaw agent" to continue refining.
$ sciclaw agent -m "Review the discussion section in manuscript.docx"
Review complete. Output: manuscript-reviewed.docx
Tracked changes (7):
• Para 2: Strengthened causal language → hedged to "these findings suggest"
• Para 3: Added missing citation for the Bhatt et al. replication study
• Para 5: Flagged unsupported claim about treatment efficacy
Comments (3):
• "Consider acknowledging the sample size limitation here"
• "This effect size is smaller than reported in the meta-analysis — worth noting"
• "The transition between paragraphs 4 and 5 could be smoother"
Open manuscript-reviewed.docx in Word to accept or reject changes.
Tracked changes (7):
• Para 2: Strengthened causal language → hedged to "these findings suggest"
• Para 3: Added missing citation for the Bhatt et al. replication study
• Para 5: Flagged unsupported claim about treatment efficacy
Comments (3):
• "Consider acknowledging the sample size limitation here"
• "This effect size is smaller than reported in the meta-analysis — worth noting"
• "The transition between paragraphs 4 and 5 could be smoother"
Open manuscript-reviewed.docx in Word to accept or reject changes.
12 built-in skills
Installed automatically when you set up sciClaw. No plugins to find or configure.
Word, PowerPoint, PDF, Excel
- Document review — tracked changes, comments, semantic diff
- Presentations — create and edit PowerPoint slides
- PDF & spreadsheets — create, merge, split, analyze
Manuscripts and figures
- Quarto authoring — loop-driven .qmd manuscripts
- Mermaid diagrams — publication-grade flowcharts
- Language polish — natural tone refinement
Evidence you can trace
- Experiment provenance — metadata capture for every run
- Benchmark logging — records with acceptance criteria
- Session history — every conversation is stored
Get started in three steps
Works on macOS, Linux, and Windows. No coding required.
1
Install
If you have Homebrew, it's one command. Binaries are also available for direct download.
brew tap drpedapati/tap && brew install sciclaw && sciclaw onboard --yes && sciclaw doctor
macOS: install Quarto for manuscript rendering with brew install --cask quarto.
2
Connect your AI provider
sciClaw works with OpenAI, Anthropic, Google Gemini, and more. Authenticate with a single command.
sciclaw auth login --provider openai
3
Chat where you already work
Recommended for non-technical users: pair Telegram/Discord and talk to sciClaw from a chat app.
sciclaw channels setup telegram
Then run sciclaw gateway. Full guide: Scientist Setup. CLI works too: sciclaw agent -m "…".
Works with your preferred AI
Switch between providers and models anytime. You bring your own API key.
OpenAI
Anthropic
Google Gemini
OpenRouter
DeepSeek
Groq
Switch models: sciclaw models set anthropic/claude-opus-4-6
Frequently asked questions
Do I need to know how to code?
No. You interact with sciClaw through plain-language conversation. If you can open a terminal and type a sentence, you can use sciClaw.
Is my data sent to the cloud?
sciClaw runs on your computer. Your conversations are sent to whichever AI provider you choose (OpenAI, Anthropic, etc.) using your own API key. Nothing is shared with us.
Can I trust AI-generated content for research?
sciClaw is designed for trust: it captures evidence trails, logs provenance metadata, and always shows its sources. You review and approve everything before it becomes part of your work.
What operating systems are supported?
macOS (Apple Silicon and Intel), Linux (amd64, arm64, riscv64), and Windows (amd64). Pre-built binaries are available on the releases page.
Is it free?
Yes. sciClaw is MIT licensed and completely free. The only cost is your AI provider's API usage, which you control directly.