Evaluation harness for LLM features in Laravel: golden datasets, code-based graders, a required falsifiability control, CI exit codes, and a Claude driver built on the official Anthropic PHP SDK.

Parse Laravel logs, cluster recurring errors and get AI-powered root cause analysis and fix suggestions.

Private manual-first document extraction with local OCR, quality scoring and targeted AI Hub fallback for Laravel.

AI agents for laravel-mutable-content apps: long-term memory and a run journal with conversation history for laravel/ai agents.

Store and query per-request AI token usage and cost from Headroom.

Token counting and context limits for the Laravel AI SDK.

Laravel integration for mtubis/html-translator: translate HTML with LLMs (OpenAI, Anthropic, Gemini, Ollama, …) while preserving markup — config, facade, Artisan, cache, logging and a test fake.

Self-hosted metering + budget enforcement for AI/LLM & agent calls in Laravel. Traces, token/cost attribution, a built-in dashboard, and multi-scope spend limits — with no external service.

Seed Eloquent models with realistic data generated by an LLM (OpenAI or any OpenAI-compatible server)

Laravel AI SDK driver for self-hosted, OpenAI-compatible vLLM servers such as the NVIDIA DGX Spark, with Cloudflare Access support.

A Laravel agent harness with swappable loops. Pick how your AI agent thinks, switch with one line, compare which works best.

CLIProxyAPI for Laravel: a laravel/ai driver over either wire protocol, gateway-first routing with a fallback of your choice, the management API, and batches run as queued calls.

Self-healing Laravel apps: turns a production exception into a proven, reviewed pull request with a failing test, a minimal patch and red-to-green evidence.

phpClaw adapter for Laravel. Add AI agents to any Laravel app with a single composer require.

Turn Livewire components and Laravel services into agent-ready tools with PHP 8 attributes. One definition powers your UI, LLM tool calling, and MCP endpoints.

Cost-governed LLM execution for Laravel: per-profile spend ceilings enforced before the call, model fallback, JSON Schema validation with one re-ask, and real cost accounting per run.

Vendor-neutral AI agent application runtime for Laravel with governed tools, deterministic workflows, connectors, MCP/A2A, knowledge, deployments, evals, and observability.

A component kit (turn/event data model + Livewire chat UI) for laravel/ai agents.

OpenTelemetry tracing for the Laravel AI SDK, following the OpenTelemetry GenAI semantic conventions.

Turn any Laravel app into a secure AI agent: automatic and custom vector embedding generation, multi-provider AI switching with auto-failover (Gemini, OpenAI/ChatGPT, DeepSeek, Anthropic/Claude, or any custom LLM), zero-config controller-to-tool discovery, and PostgreSQL/MySQL/MongoDB/Qdrant/Pinecone vector database support.

IntentQL turns natural-language prompts into safe, validated Laravel database queries. The language model plans; it never writes SQL.

Laravel package for on-device / local language models (MLX, Ollama, LM Studio, llama.cpp, vLLM)

Define AI-agent behavior as test cases, run them against your real agent, and automatically catch regressions before they ship.

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.

Premium multi-provider Laravel AI SDK — AIHub::fake() testing engine, multi-turn chat sessions, SSE streaming, structured JSON schemas, Gemini 3.8, Claude 3.7 thinking, GPT-5.6/o3, 13 providers, Studio playground, failover, and cost tracking.

Production RAG for Laravel: pgvector retrieval with citations, and an eval harness that fails CI when retrieval quality regresses.

Production AI governance layer for Laravel — prompt management, model routing, cost tracking, rate limiting, caching, audit logging, and tenant quotas on top of the official Laravel AI SDK.

Durable, tenant-aware AI workflow compilation and execution for Laravel

Pandora — a Laravel-native agentic framework: agents, tools, approvals, automations, memory and a Livewire control center.

Reusable AI chatbot package for Laravel with polymorphic module attachment, context window management, and guest/authenticated token quotas.

Shared, provenance-aware AI pricing and cost attribution for Laravel applications.

List and cache the currently-available models for each provider configured in the Laravel AI SDK.

Conversation / turn / proposal runtime for the Laravel Capabilities bus — pluggable LlmClient, progress store, bus-only tools

A unified, production-minded LLM client for Laravel — Claude and OpenAI behind one interface, with streaming, tool calling, Blade prompt templates, response caching, retries/fallbacks and per-request cost tracking.

Add AI-powered computed attributes to any Eloquent model — a thin Eloquent layer on top of the official Laravel AI SDK.

Official Laravel integration for LatentKit's route-based AI gateway.

Turn Laravel exceptions into dev-friendly explanations and user-safe messages using your own LLM keys. Multi-provider, multi-language.

Agentic AI layer for padosoft/laravel-flow: LLM prompt nodes, MCP client and server (expose flows as MCP tools), bounded agent nodes, AI flow builder and Flow Advisor. Part of the Laravel Flow 2.0 suite.

Agent Discovery Protocol implementation for Laravel APIs built on top of rest-generic-class metadata.

AI backend for Laravel: provider/model catalog with pricing, request usage tracking and cost reporting, agent system-prompt registry and a self-updating tariff registry.

Unified translation services (DeepL, Google Cloud Translation, ...) for Laravel.

An abuse firewall for AI endpoints: detect runaway agent loops within a request, and open a circuit on a per-tenant spend anomaly, before a self-recursing agent drains the API budget. Security-framed, not just rate limiting.

Static analysis for insecure AI wiring in Laravel apps: a scan command that finds AI provider keys leaked into tracked or client-reachable files, plus PHPStan rules that flag model output flowing into unsafe sinks and prompts built by concatenation. SARIF for CI.

Scan, redact, and optionally block outbound requests to AI providers (OpenAI, Anthropic, Gemini, and more) for leaked secrets and PII. Review in an inbox and fail CI before a prompt leaks. SDK-agnostic, at the HTTP layer.

Cost reporting and budget guardrails for the Laravel AI SDK.

Server-side AI crawler & bot traffic tracking for Laravel, reporting to DataFast. Unofficial PHP port of @datafast/ai-crawl.

A safety layer for LLM input and output in Laravel: a fluent, composable pipeline of guards (PII redaction, prompt-injection detection, moderation, schema validation) plus a provider-agnostic wrapper for any LLM call.

A deep RAG search orchestrator for Laravel, based on Bidirectional Evolutionary Search (BES): backward goal decomposition, evolving evidence trails and cited, auditable answers.

DBMind 1.0 — the SQL-only edition: a local Ollama model writes guarded, read-only SQL (with JOINs) against your approved tables and answers from the real rows. No Qdrant, no embeddings.

First-class Laravel integration for LLMesh — the flexible PHP LLM SDK