CellType CLI — An autonomous agent for drug discovery research
Project description
celltype-cli
v0.2.0 is out — Offload GPU-accelerated tools to CellType Cloud with no additional setup. ESMFold, Boltz-2, RFdiffusion, ProteinMPNN, and more. Update now:
curl -fsSL https://raw.githubusercontent.com/celltype/celltype-agent/main/install.sh | bash
An autonomous agent for drug discovery research. Like Claude Code, but for biology.
Ask questions in natural language. celltype-cli plans the analysis, selects the right tools, executes them, validates results, and returns data-backed conclusions.
Benchmark
CellType CLI achieves 90% accuracy on BixBench-Verified-50, outperforming all existing agentic systems for computational biology.
| System | Accuracy |
|---|---|
| celltype-cli (Opus 4.6) | 90.0% |
| Phylo BiomniLab | 88.7% |
| Edison Analysis | 78.0% |
| Claude Code (Opus 4.6) | 65.3% |
| OpenAI Agents SDK (GPT 5.2) | 61.3% |
Why ct?
- 190+ drug discovery tools — Target prioritization, compound profiling, dose-response modeling, pathway enrichment, safety assessment, clinical development, and more.
- Claude-powered reasoning — Built on the Claude Agent SDK. Claude plans multi-step research workflows, calls tools, self-corrects, and synthesizes findings.
- Managed data pipelines — One-command download of DepMap, PRISM, L1000, and proteomics datasets with auto-discovery loaders.
- 30+ database APIs — PubMed, ChEMBL, UniProt, Open Targets, ClinicalTrials.gov, Reactome, PDBe, and more — no setup required.
- Research UX — Interactive terminal with @mentions, session resume, trace logging, HTML reports, and notebook export.
- Persistent sandbox — Stateful Python and R execution across turns (pandas, scipy, gseapy, pydeseq2, BioPython).
- Open source — MIT licensed.
Installation
Quick install
curl -fsSL https://raw.githubusercontent.com/celltype/celltype-agent/main/install.sh | bash
Detects Python 3.10+, installs the latest celltype-cli release from PyPI via uv, pipx, or pip, and launches the interactive setup wizard.
Manual install
# With pipx (isolated environment, recommended)
pipx install celltype-cli
# Or with pip
pip install celltype-cli
# Or with optional scientific stacks (RDKit, scanpy, torch, etc.)
pip install "celltype-cli[all]"
# Run the setup wizard
ct setup
Authentication
# Interactive setup (recommended)
ct setup
# Or set directly
export ANTHROPIC_API_KEY="sk-ant-..."
# Non-interactive (CI/scripting)
ct setup --api-key sk-ant-api03-...
Getting Started
Basic usage
# Start interactive session
ct
# Single query
ct "What are the top degradation targets for this compound?"
# Validate setup
ct doctor
Interactive commands
Inside ct interactive mode:
/help— command reference + examples/tools— list all tools with status/agents N <query>— run N parallel research agents/case-study <id>— curated multi-agent case studies/sessions,/resume— session lifecycle/copy,/export— save/share outputs/usage— token and cost tracking
Quick examples
Target prioritization
$ ct "I have a CRBN molecular glue. Proteomics shows it degrades
IKZF1, GSPT1, and CK1α. Which target should I prioritize?"
Combination strategy
$ ct "My lead compound is immune-cold. What combination strategy should I use?"
Resistance biomarkers
$ ct "Build a resistance biomarker panel for my lead compound"
Key Features
190+ Domain Tools
| Category | Examples |
|---|---|
| Target | Neosubstrate scoring, degron prediction, co-essentiality networks |
| Chemistry | SAR analysis, fingerprint similarity, scaffold clustering |
| Expression | L1000 signatures, pathway enrichment, TF activity, immune scoring |
| Viability | Dose-response modeling, PRISM screening, therapeutic windows |
| Biomarker | Mutation sensitivity, resistance profiling, dependency validation |
| Clinical | Indication mapping, population sizing, TCGA stratification |
| Safety | Anti-target flagging, multi-modal profiling, SALL4 risk |
| Literature | PubMed, OpenAlex, ChEMBL search |
| DNA | ORF finding, codon optimization, primer design, Golden Gate/Gibson assembly |
| Data APIs | MyGene, UniProt, Reactome, PDBe, ClinicalTrials.gov, and 25+ more |
Data Management
# Download core datasets
ct data pull depmap # DepMap CRISPR, mutations, expression
ct data pull prism # PRISM cell viability
ct data pull msigdb # Gene sets
ct data pull alphafold # Protein structures (on-demand)
# Or point to existing data
ct config set data.depmap /path/to/depmap/
ct config set data.prism /path/to/prism/
Without local data, ct still works using 30+ database APIs.
Reports
Every query auto-saves a markdown report. Convert to branded, self-contained HTML:
ct report list # list reports
ct report publish # convert latest .md to .html
ct report show # open in browser
Dark theme, responsive layout, inline CSS. No CDN, no JavaScript. Shareable via email/Slack.
Known Limitations
- Local data for some tools — Target, viability, expression, combination, biomarker tools need local DepMap/PRISM/L1000 datasets. The planner auto-excludes unavailable tools and uses API alternatives.
- Optional dependencies — RDKit (chemistry), scanpy (single-cell), torch (protein embeddings). Tools report install instructions if missing.
- API rate limits — PubMed, UniProt, Open Targets may rate-limit heavy usage.
Troubleshooting
| Symptom | Fix |
|---|---|
ct fails at startup |
ct doctor |
| No API key | ct setup or export ANTHROPIC_API_KEY=... |
| Data not found | ct data pull <dataset> |
| Missing dependency | pip install "celltype-cli[all]" |
| Session lost | ct --continue (sessions auto-save) |
Contributing
git clone https://github.com/celltype/cli.git
cd cli
pip install -e ".[dev]"
ct setup
pytest tests/
We welcome contributions — bug reports, feature requests, documentation improvements, and pull requests.
Enterprise & On-Premise
celltype-cli is free and open source, powered by Claude out of the box.
For pharma and biotech teams that need to keep data on-premise, CellType offers proprietary research agent models purpose-built to replace frontier LLMs — deployable locally behind your firewall with zero data leaving your infrastructure.
Contact us to learn more.
License
MIT — see LICENSE
Built by CellType Inc.
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file celltype_cli-0.2.1.tar.gz.
File metadata
- Download URL: celltype_cli-0.2.1.tar.gz
- Upload date:
- Size: 1.6 MB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.10.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
2b19b2c1d7a92ca7ba6948b3a5d92bf2112c3e3c56f8e32dd6f5b85d10ff1f3a
|
|
| MD5 |
e10ded1ba47615b1b79180036d89d5de
|
|
| BLAKE2b-256 |
2410be01aa7b59677195afe50727d85e948c20a526021d4ad7b11da89503c608
|
File details
Details for the file celltype_cli-0.2.1-py3-none-any.whl.
File metadata
- Download URL: celltype_cli-0.2.1-py3-none-any.whl
- Upload date:
- Size: 545.8 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.10.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
cd6d541c63d02eaf76bc91adbf6f2675d3a0eb4a39ee2eb5b726b7ec4bbc2e5a
|
|
| MD5 |
b8197f68bda937df5d1e61cca0df2e6b
|
|
| BLAKE2b-256 |
a6b03525a54e0413da031771b5aebc244ad8ae7859df36dbb907c7a391ec8bc7
|