Modularity and Collaborative Development
Leverage tools, too libraries, and interfaces
12 articles
- Modular reuse of toolsShare and reuse AI tools, content collections, and workflow logic
- Building thunks from reusable partsHow tools, tool libraries, interfaces, and MCP exports let you compose thunks from parts that can be built, tested, and shared on their own.
- Tool call history and performanceSee every call to a tool, what it was given and returned, and for AI tools what each call cost and how long it took.
- Custom ToolsA guide to building and configuring custom tools in the thunk builder, including AI-based, code-based, and other tool types.
- How custom code tools workThe JavaScript format, available globals, and how a custom code tool reads inputs, calls other tools, and returns a result.
- How to build a Custom Database toolGive the AI agent a tool that queries your SQL database, so a workflow can look up or read records straight from the source.
- How to build a Spreadsheet Lookup toolGive the AI agent a tool that looks up rows in a Google Sheet, Excel workbook, or CSV by filtering on the columns you choose.
- How to build a Tool WrapperTurn an existing tool into a new, purpose-built one by fixing some of its arguments and giving it a name and inputs of your own.
- Promote _meta arguments on MCP toolsAdvertise extra MCP _meta settings as ordinary tool arguments so the agent can fill them in.
- Publish your thunk as an MCP server for AI agentsPublish a thunk's tools as an MCP server so Claude or another AI agent can connect and call them
- Reuse thunk interfaces and plug-in connectionsExport a thunk's tool signatures as an interface, require that interface as a plug-in connection, or create an MCP server that implements it
- Automated tests for custom toolsGive each custom tool its own test cases: author them by hand or with AI, capture them from real calls, and re-run the suite as the tool changes.