Why
Large containers can flood an agent's context and burn RUs. Server-side spikes (~5k req/day) suggest automation already hits large result sets. Agents need explicit, bounded, resumable paging.
Proposed behavior
- Query/ls tool results always return a continuation token and honor a max-items cap; agents can resume explicitly.
- Enforce a safe default cap for MCP callers.
Acceptance criteria
- Continuation token surfaced in structured output.
- Default cap applied for MCP.
- Documented.
- Tests.
Filed from the Agentic & Automation Roadmap (docs/agentic-roadmap.md), item M6, Wave 2. Priority P1.
Why
Large containers can flood an agent's context and burn RUs. Server-side spikes (~5k req/day) suggest automation already hits large result sets. Agents need explicit, bounded, resumable paging.
Proposed behavior
Acceptance criteria
Filed from the Agentic & Automation Roadmap (
docs/agentic-roadmap.md), item M6, Wave 2. Priority P1.