Full Documentation · uv pip install scitex-hub[all]
| # | Problem | Solution |
|---|---|---|
| 1 | Fragmented tools. Literature, writing, analysis, and visualization require separate, often proprietary applications, forcing constant context-switching and making it difficult for AI agents to build sufficient context across the research workflow. | Unified platform. Scholar, Writer, FigRecipe, Console, Hub, and Clew in a single Django web application, deployable anywhere with Docker. All apps share the same project filesystem and integrate through the scitex Python package. |
| 2 | No custom tooling. Every research group needs domain-specific tools (e.g., clinical trial dashboards, spike-sorting interfaces, compound screening pipelines), yet building and sharing them requires deep computational knowledge and creating components from scratch. | App Maker and Store. Researchers create, publish, and install custom research tools on top of shared components — user/group permissions, AI infrastructure, containerized computation, and file operations are handled by the platform. |
| 3 | AI tools not research-aware. Existing tools often lack AI assistant capabilities and domain-specific skills for scientific work, unable to operate across the full research lifecycle (literature review, analysis, writing, verification). | Built-in AI co-pilot. Platform-aware context, skills, and tools such as MCP (Model Context Protocol) and CLI span the full research lifecycle, providing an AI assistant that understands the entire project from natural language. |
| 4 | Review crisis. The growing volume and heterogeneity of published papers overwhelms a limited, volunteer-based peer review process that cannot scale. | Open review via Issues and PRs. GitHub-style issue tracking and pull requests bring transparent, structured, and scalable peer review to research projects — anyone can inspect, comment, and propose changes. |
| 5 | Broken provenance. Papers, code, and execution environments are rarely tied together, making it difficult for reviewers to verify claims and for other researchers to replicate results — slowing cumulative scientific progress. | Verifiable provenance. Clew links papers, code, data, and execution environments into a hash-verified DAG (Directed Acyclic Graph) with visualization that serves as a compressed view of the research workflow and logic — reducing the decision points reviewers must check. |
| 6 | Lost knowledge on handoff. When researchers graduate or leave a project, successors inherit scattered files with little context, making it difficult to understand where to pick up and continue the work. | Seamless project handoff. The full project state — code, data, provenance graph, manuscript drafts, and execution environment — lives in one place, so successors can understand and continue work immediately. |
| 7 | No research community platform. No GitHub-like infrastructure exists for research-project-centric, fully traceable, parallel-working collaboration. | GitHub-style project hub. Repository hosting and ticket-based development with co-authors and the community enable efficient research advancement and collaboration. |
| 8 | No control. Researchers have no ownership over their infrastructure: vendor lock-in, opaque algorithms, unilateral pricing changes, and data policies they cannot influence. | Self-hosted, open-source, runnable from anywhere. Deploy on your laptop, lab server, or cloud. AGPL-3.0 licensed — inspect every line, customize freely, no vendor lock-in, no data surrender. |
Table 1. Eight infrastructure challenges in scientific research and how SciTeX Hub addresses each. These gaps fuel the reproducibility crisis, limit what AI can do for research, and leave knowledge stranded when people move on.
SciTeX Hub is an AI-native infrastructure so that researchers can focus on science, not on tooling.
Writer
Scholar
Apps
Figure 1. Core application modules. Writer provides a LaTeX manuscript environment with live compilation. Scholar offers literature discovery, BibTeX enrichment, and PDF management. The Apps panel shows the project-centric hub linking all modules.
graph TB
subgraph workspace[apps/workspace]
S[scholar_app] --- W[writer_app]
W --- F[figrecipe / plt]
F --- CN[console_app]
CN --- H[repo_app]
H --- CW[clew_app]
end
subgraph infra[apps/infra]
WS[workspace_app] --- PA[platform_app<br/>DataStore / FileVault / JobQueue]
PA --- PR[project_app]
PR --- A2[a2a_app]
end
workspace --> infra
infra --> DK[Docker / Postgres / Gitea]
scitex-hub/
├── apps/
│ ├── workspace/ # scholar / writer / figrecipe / console / hub / clew
│ ├── infra/ # workspace_app, platform_app, project_app, a2a_app
│ └── public_app/ # landing page + public tools
├── deployment/docker/ # docker_dev / docker_prod / envs
├── config/ # Django settings
├── src/scitex_hub/ # pip package: CLI + MCP server
└── tests/
pip install scitex-hub # CLI only
pip install scitex-hub[mcp] # CLI + MCP server
pip install scitex-hub[all] # Everythinggit clone https://github.com/ywatanabe1989/scitex-hub.git
cd scitex-hub
make start # Start development environment
# Access at: http://localhost:8000
# Gitea: http://localhost:3000
# Test user: test-user / Password123!Python API
import scitex_hub
# Version and health
scitex_hub.__version__ # read from pyproject.toml (e.g. "0.17.0-alpha")
scitex_hub.get_version() # Version string
scitex_hub.health_check() # Local package info
scitex_hub.health_check("https://scitex.ai/api/health/") # Remote endpoint
# Clients / helpers
client = scitex_hub.CloudClient() # HTTP client
env = scitex_hub.get_environment() # Environment config
docker = scitex_hub.DockerManager() # Container helpersCLI Commands
scitex-hub --help # Help
scitex-hub --help-recursive # All commands recursively
scitex-hub --version # Version
# Git hosting (Gitea)
scitex-hub gitea list # List repositories
scitex-hub gitea clone user/repo # Clone repository
scitex-hub gitea push # Push changes
scitex-hub gitea pr create # Create pull request
scitex-hub gitea issue create # Create issue
# Docker management
scitex-hub docker up # Start containers
scitex-hub docker down # Stop containers
scitex-hub docker ps # Container status
scitex-hub docker build # Build images
scitex-hub docker restart # Restart services
# MCP server
scitex-hub mcp start # Start MCP server
scitex-hub mcp list-tools # List available tools
scitex-hub mcp doctor # Diagnose setup
scitex-hub mcp installation # Client config instructions
# Utilities
scitex-hub status # Deployment status
scitex-hub completion # Shell completion setup
scitex-hub list-python-apis # List all Python APIsMCP Server — for AI Agents
AI agents can interact with the SciTeX Hub platform autonomously via MCP (Model Context Protocol) tools.
| Category | Tools | Description |
|---|---|---|
| gitea | 14 | Git operations (clone, push, pull, PR, issues, auth) |
| sdk | 14 | DataStore, FileVault, JobQueue operations |
| api | 9 | Scholar search, CrossRef, BibTeX enrichment |
| app | 7 | App plugin lifecycle (init, validate, submit) |
| onsite | 6 | On-site platform operations |
| project_crud | 5 | Project create, list, rename, delete |
Table 2. MCP tool categories — 55 tools total registered via
register_all_tools in
_mcp_tools/init.py. Use scitex-hub mcp list-tools
for the live list.
Claude Desktop (~/.config/claude/claude_desktop_config.json):
{
"mcpServers": {
"scitex-hub": {
"command": "scitex-hub",
"args": ["mcp", "start"]
}
}
}Skills — for AI Agents
Skill files provide context-aware guidance to AI agents working within the SciTeX ecosystem.
# Export skills to dotfiles (sync to Claude)
scitex-dev skills export --package scitex-hub
# List available skills
scitex-hub skills listSkills are stored in src/scitex_hub/_skills/scitex-hub/ and cover deployment, development, testing, and more.
Deployment
make start # Development (default)
make ENV=prod start # Production
make ENV=prod status # Health check
make ENV=prod db-backup # Backup database
make help # All available commandsConfiguration
.env files in deployment/docker/envs/ (gitignored):
.env.dev # Development
.env.prod # Production
.env.staging # Staging
.env.example # Template (tracked)Key variables:
SCITEX_HUB_DJANGO_SECRET_KEY=your-secret-key
SCITEX_HUB_POSTGRES_PASSWORD=strong-password
SCITEX_HUB_GITEA_TOKEN=your-tokenProject Structure
scitex-hub/
├── apps/ # Django applications
│ ├── workspace/ # Workspace modules
│ │ ├── apps_app/ # App marketplace & dev install
│ │ ├── scholar_app/ # Literature discovery
│ │ ├── writer_app/ # Scientific writing
│ │ ├── console_app/ # Terminal & code execution
│ │ ├── repo_app/ # Project hub & file browser
│ │ └── clew_app/ # Verification pipeline
│ ├── infra/ # Platform infrastructure
│ │ ├── workspace_app/ # Module registry & workspace shell
│ │ ├── platform_app/ # DataStore, FileVault, JobQueue APIs
│ │ └── project_app/ # Project management
│ └── public_app/ # Landing page & public tools
│
├── deployment/docker/
│ ├── docker_dev/ # Development compose
│ ├── docker_prod/ # Production compose
│ └── envs/ # .env files (gitignored)
│
├── config/ # Django settings
├── static/ # Shared frontend assets
├── src/scitex_hub/ # pip package (platform CLI + MCP)
├── tests/ # Test suite
└── Makefile # Thin dispatcher
For app developers: Use
pip install scitex-app[cli]and thescitex-app appCLI. scitex-hub is the platform server — app developers don't need to install it.
scitex-hub is part of SciTeX. Install via
the umbrella with pip install scitex[hub] to use as
scitex.hub (Python) or scitex hub ... (CLI).
| From | Produces | To | Outcome |
|---|---|---|---|
| Scholar | Citations as cards | Writer | Convenient, evidence-based referencing |
| SciTeX-followed Analysis | Artifacts | Writer | AI writes a manuscript based on actual results |
| FigRecipe | Style-editable, composable figures | Writer | Publication-ready figures in context |
| Clew | Verification and DAG visualization | Writer | Proven reproducibility for every claim |
The SciTeX system follows the Four Freedoms for Research below, inspired by the Free Software Definition:
Four Freedoms for Research
- The freedom to run your research anywhere — your machine, your terms.
- The freedom to study how every step works — from raw data to final manuscript.
- The freedom to redistribute your workflows, not just your papers.
- The freedom to modify any module and share improvements with the community.
AGPL-3.0 — because we believe research infrastructure deserves the same freedoms as the software it runs on.
scitex-hub serves the Google A2A protocol at a2a.scitex.ai for the orochi agent fleet — AgentCard discovery, JSON-RPC dispatch, bearer-auth via Gitea PAT, and a Tier 3 forwarder to live agents. See apps/infra/a2a_app/README.md.
curl https://a2a.scitex.ai/v1/agents/ | jq '.agents | length'SciTeX Hub is in alpha. Core functionality is working and under active development. Data formats may change between releases — back up important work.
We welcome contributions! See CONTRIBUTING.md.
