arXiv:2606.19992v1 Announce Type: cross Abstract: In the agentic web era, LLM-based agents increasingly invoke web services as tools, yet most interfaces remain \emph{static endpoints} that poorly express long-horizon workflows with loops, conditionals, joins, and retries. We present ToolPro, which represents an agent's tool intent as an \emph{executable tool program} that compactly encodes multi-step service interactions with explicit effect types. ToolPro combines constraint-guided program construction, effect-aware replay for exactly-once state-modifying calls, and a profile-driven policy t
The rapid advancement and adoption of LLM-based agents necessitate more sophisticated interfacing for complex web service interactions.
Improved agent-tool interaction is crucial for realizing the full potential of autonomous AI agents in automating multi-step workflows.
Current static endpoint limitations are addressed by a programmatic approach, enabling agents to execute more complex, long-horizon tasks with web services.
- · AI Agent Developers
- · SaaS Providers (with sophisticated APIs)
- · Enterprises Adopting AI Automation
- · Providers of Simple API Gateways
- · Businesses reliant on static, manual web service integration
AI agents can now perform more complex, multi-step operations on the web autonomously.
This could lead to a significant increase in the types and sophistication of tasks agents can automate, further collapsing certain white-collar workflows.
The enhanced programmability of agent interactions might accelerate the development of entirely new categories of agentic web services and business models.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI