Expose Eloquent models to an AI agent without handing it your database. Field allowlists, per-record policy checks, result ceilings and an audit trail, all failing closed.
Laravel Agent Kit exposes three read-only tools (list, get, search) with clear naming and basic descriptions. All tools have input schemas with type definitions and parameter descriptions. However, descriptions are generic and lack LLM-optimized guidance on when to use each tool vs. alternatives. Output schemas are not documented. No error handling guidance, no examples of what fields are returned, and no documentation of pagination behavior or result limits despite mentioning 'maxResults' in parameter descriptions. Security model is sound (field allowlists, per-record policy checks) but not surfaced in tool descriptions. Tool composition is clean (single responsibility), but descriptions don't explain the relationship between list/search or when to prefer one.
Get a single record by identifier with optional relation expansion
List records with optional filtering, pagination, and relation expansion
Search records by free text with optional pagination and relation expansion
Output schemas not documented. LLMs cannot infer what fields are returned or plan downstream tool calls. No documentation of which fields are included in responses or how relations are structured.
Descriptions lack LLM-optimized guidance. Generic descriptions ('List records...', 'Get a single record...') do not explain WHEN to use each tool or how they differ. No guidance on pagination, result limits, or performance implications.
Pagination and result limits mentioned in parameter descriptions but not in tool descriptions. LLMs need explicit guidance: 'Results capped at resource maxResults setting' should be in the main description, not buried in a parameter note.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 62 | 2026-07-28+ | v2 |
No error handling guidance. Tool descriptions do not explain what errors are possible, when they occur, or what the LLM should do next (e.g., 'If authorization fails, the agent cannot access this record').
Security model (field allowlists, per-record policy checks) is a key differentiator but not mentioned in tool descriptions. LLMs should understand that results are filtered by authorization, not that all fields are exposed.