LLM-powered desktop lab for Python, R, and 80 bioinformatics DBs

What it does Wisp Science is a cross-platform desktop app (and headless CLI) that connects OpenAI-compatible and Anthropic models to persistent Python and R environments. It bundles roughly 80 bioinformatics and computational biology database clients via MCP servers, and it can execute code locally, over SSH, in WSL, or on GPU runtimes. Conversations and project state live in a local SQLite database, while API keys and passwords stay in the OS keyring.

The interesting bit The architecture is aggressively modular—ten separate Rust crates handle everything from LLM streaming to SQLite compaction—yet the product focus is on scientific reproducibility. It uses plain-text SKILL.md files for reusable agent workflows, probes remote compute before trusting it, and isolates runtime selection per conversation so a misbehaving SSH host can’t silently derail your session.

Key highlights

• Local-first by design: models, code, and data stay on your machine or explicitly chosen remote servers.
• Bundled MCP bioinformatics servers provide access to around 80 databases out of the box.
• Supports local, SSH, WSL, and GPU compute runtimes with per-conversation resource isolation.
• Rust/Tauri v2/Leptos stack with a headless CLI mode for server or terminal use.
• Secrets live in the OS keyring, never in the SQLite store or on remote hosts.

Caveats

• Explicitly an MVP vertical slice; the README notes deferred features on the roadmap.
• Linux is currently source-build only; Windows and macOS have prebuilt installers.
• Optional R support requires manual setup (jsonlite must already be installed); Wisp does not auto-provision R packages.

Verdict Bioinformaticians and computational biologists tired of cloud notebooks should take a look. If you don’t need LLM-assisted Python/R workflows or local bio-database access, this is overkill; and if you need a polished, feature-complete product today, the MVP status may frustrate you.