{"id":"ywatanabe1989-scitex-python","name":"scitex-python","homepage":"https://scitex.ai","repo_url":"https://github.com/ywatanabe1989/scitex-python","category":"ai-ml","subcategories":[],"tags":["ai-ml","reproducibility","scientific-computing","mcp","research-automation","statistics","latex","data-visualization"],"what_it_does":"scitex (SciTeX) is a Python toolkit and orchestration layer for scientific research workflows: experiment tracking/session management, unified file I/O, reproducible figure generation (via figrecipe), statistical testing, literature management (scholar), LaTeX manuscript compilation (writer), and cryptographic verification/provenance tracking (clew). It also exposes an MCP server surface (named scitex) with a large set of MCP tools intended for AI agents to run parts of the research lifecycle via structured tool calls.","use_cases":["Reproducible data analysis pipelines with logging, frozen configs, and deterministic seeds","Generating publication-style figures with data+recipe exports for later re-rendering","Running statistical hypothesis tests and formatting results for papers","Automated literature search/fetch and BibTeX enrichment","Compiling LaTeX manuscripts from figures/tables/csv inputs","Cryptographically verifying that generated outputs/manuscript claims match source data via hash-chain DAGs","Using MCP-enabled tool calls to let AI agents perform research steps (search, stats, figures, writing) in a structured way","Local/Script-friendly automation with CLI commands wrapping key modules"],"not_for":["A fully hosted SaaS API that provides stable network endpoints (this is primarily a local Python/MCP/CLI toolkit)","Security-critical deployments without verifying/controlling the underlying network integrations (e.g., LLM providers, web automation, dataset sources)","Use cases that require formal REST/OpenAPI contracts (documentation for REST/OpenAPI is not evidenced here)","Production environments needing guaranteed SLAs for uptime"],"best_when":"You want an agent-friendly, modular research automation stack in Python where outputs are tracked and can be verified/provenanced locally, and you can run the MCP server/tooling yourself.","avoid_when":"You need strict, formally specified network API contracts (REST/OpenAPI) and comprehensive documented operational guarantees (SLA, rate limits, retry semantics) for a hosted service.","alternatives":["Other scientific workflow/orchestration tools (e.g., Airflow, Prefect) paired with notebook-based analysis libraries","Reproducibility tooling like DVC/LFS for data+artifact versioning","Agent tool frameworks (LangChain/LangGraph) with custom tools for stats/figures rather than an integrated MCP suite","General-purpose bibliography managers and scripting (Zotero + Python scripts, Semantic Scholar APIs, Crossref/OpenAlex clients)"],"af_score":57.2,"security_score":51.5,"reliability_score":33.8,"package_type":"mcp_server","discovery_source":["github"],"priority":"high","status":"evaluated","version_evaluated":null,"last_evaluated":"2026-03-30T15:18:49.602435+00:00","interface":{"has_rest_api":false,"has_graphql":false,"has_grpc":false,"has_mcp_server":true,"mcp_server_url":null,"has_sdk":true,"sdk_languages":["python"],"openapi_spec_url":null,"webhooks":false},"auth":{"methods":["Environment/config-based credentials for optional integrations (e.g., LLM providers, web automation, cloud integrations) are implied but not specified in the provided content"],"oauth":false,"scopes":false,"notes":"No explicit authentication scheme (API keys, OAuth, scopes) is documented in the provided README excerpt for the core scitex toolkit/MCP server. Authentication likely depends on the configured optional modules/integrations (e.g., LLM backends), but this is not specified here."},"pricing":{"model":null,"free_tier_exists":false,"free_tier_limits":null,"paid_tiers":[],"requires_credit_card":false,"estimated_workload_costs":null,"notes":"Open-source Python package (AGPL-3.0 per provided manifest). Costs are primarily compute and any third-party service usage (e.g., LLM APIs) rather than a platform pricing tier."},"requirements":{"requires_signup":false,"requires_credit_card":false,"domain_verification":false,"data_residency":[],"compliance":[],"min_contract":null},"agent_readiness":{"af_score":57.2,"security_score":51.5,"reliability_score":33.8,"mcp_server_quality":75.0,"documentation_accuracy":70.0,"error_message_quality":0.0,"error_message_notes":null,"auth_complexity":60.0,"rate_limit_clarity":10.0,"tls_enforcement":85.0,"auth_strength":40.0,"scope_granularity":30.0,"dependency_hygiene":50.0,"secret_handling":55.0,"security_notes":"Core README highlights reproducibility and cryptographic verification (clew) but does not document security controls for authentication/authorization, least-privilege scopes, or secrets management. Optional modules likely interact with third-party services (LLM providers, web automation, datasets), so operational security depends on environment configuration, network TLS usage, and how those modules handle credentials and logging. The package includes an audit module flag in optional dependencies, but its availability/coverage is not shown in the excerpt.","uptime_documented":0.0,"version_stability":55.0,"breaking_changes_history":45.0,"error_recovery":35.0,"idempotency_support":"false","idempotency_notes":null,"pagination_style":"none","retry_guidance_documented":false,"known_agent_gotchas":["MCP tool surface is very large (293 tools); agents may require careful tool selection/guardrails to avoid unintended long-running or external-network tasks (scholar fetch, browser automation, dataset downloads).","The toolkit mixes local file operations, external fetches, and optional integrations; agent planners should model side effects and artifacts (saved files, generated figures, compiled outputs).","No explicit retry/backoff or idempotency guidance is shown in the provided excerpt, so agents may need conservative re-run strategies."]}}