Virtual Tool Calling and the Future of Portable AI Toolchains
summary
Claim: translating tool definitions into structured system message tools lets any model invoke tools without provider APIs. Demo shows a JS calculator and a percent-button bug fixed via a remote agent flow on the continue platform, with injected XML in the system message and a local run on Gemma 4B that actually triggers a search tool. Workflows extend to VS Code, a CLI, and Sentry integration that can create GitHub issues or PRs. Result: one codebase runs across Claude, OpenAI/GPT-5, local models and hosted providers, with some UI glitches but successful tool invocation.
AI-written summary · about
chapters (8)
- 0:00Introduction and setup
- 3:19Speaker background and Continue
- 5:48Virtual tool calling explained
- 9:01Demo: Calculator app and bug fix
- 13:54Portability of tool calling
- 17:46Legacy dev tools and MCP
- 20:39Demo: Continuous AI workflow
- 24:29Amplified dev and closing
related talks
01h 01mAI Agents In-Depth – Function Calling, MCP and Tool Use Under the Hood
30mBuilding Tools with Tools: Modern AI-Driven Development
56mBeyond Chat: Bringing Models to the Canvas
32mProgramming language evolution in the AI era
29mPairing with the machine: How AI Agents are rewriting the rules of software development
54mKeynote: AI-Powered App Development
30mDeveloping AI Integration in Domain Specific Tools with Theia AI
29mTools and Roadmap for AI-Native Developers