Tools and pipelines you can download, drop in, and compose.
Everything here is a LazyTool. That is the point.
Four tools are available now. More are in active development — listed below as a preview of what is coming.
The official LazyBridge Claude Code skill. Gives any Claude Code session native understanding of the full LazyBridge API — every class, pattern, and rule. No docs to paste, no API to explain.
Reads PDF, DOCX, HTML, MD, and TXT files from a folder or single path. Returns LLM-ready text with file headers. Degrades gracefully if optional dependencies are missing.
Turn any documentation folder into a queryable local skill. Index once with build_skill(), then attach the skill to any agent via skill_tool(). BM25 retrieval — no embeddings, no vector DB, no API calls.
Two-agent research chain: researcher → writer via LazyContext. Includes a mock search tool and typed Report output (Pydantic). Multi-provider. Side-by-side comparison with the raw SDK included.
Four production-ready variants of the same investment research workflow — orchestrator-directed, nested pipeline, fan-out synthesis, and fully typed multi-provider. Pick the architecture that fits.
Fan-out web search across multiple queries in parallel via sess.as_tool(mode="parallel"). Results concatenated and returned as a single tool call. Drop into any agent loop.
Connects to SQL databases (SQLite, PostgreSQL, MySQL) and exposes a natural-language query interface as a LazyTool. The agent describes what it needs; the tool translates and executes.
Multi-agent risk assessment pipeline. Parallel specialists (market, credit, liquidity, operational) feed a risk synthesizer that produces a structured report with severity ratings and recommended actions.
No items match this filter.
Want to contribute?
The toolbox grows when people share. Open a PR on GitHub with a new tool or pipeline.