Nova Mcp Laravel Package

Technical Evaluation

Architecture fit: opscale-co/nova-mcp is a Laravel package that bridges Laravel Nova with the Model Context Protocol (MCP), exposing Nova resources as MCP tools and (as of v1.2.0) prompts. It fits naturally into Laravel’s package ecosystem via a service provider and configuration file. The new promptsresolver feature extends the MCP surface area without altering the core tool‑exposure architecture, suggesting a modular design where prompts are resolved through a dedicated component. This aligns with MCP’s separation of tools, prompts, and resources.

Integration feasibility: Installation is standard Composer (composer require opscale-co/nova-mcp), followed by publishing config and registering the service provider. The package likely exposes an HTTP endpoint for MCP clients (e.g., Claude Desktop, Cursor, custom agents). The prompts resolver adds the ability to load predefined prompt templates, which may require configuring a prompt directory or resolver class. Feasibility is high for teams already on Laravel Nova, but requires an MCP‑capable client and an understanding of MCP’s prompt semantics. The merge commit in the release notes is housekeeping and does not affect integration.

Technical risk: The package is at v1.2.0 with a small but active commit history. Risks include: (1) MCP specification evolution – prompts are a newer MCP primitive, and the resolver’s API may change; (2) security – prompts can carry instructions that influence LLM behaviour, so resolver input validation and authorization are critical; (3) sparse release notes – the exact resolver contract (interface, discovery mechanism, caching) is undocumented in the notes; (4) the merge commit suggests branch hygiene issues that could hint at process maturity concerns. Overall risk is moderate, suitable for internal or low‑risk AI assistant integrations.

Key questions:

• What is the exact interface for the prompts resolver? How are prompts discovered (filesystem, database, config)?
• Does the resolver enforce Nova authorization policies for prompts, or are prompts globally available?
• Are there version constraints for Laravel, Nova, and PHP? Is Nova 4 required?
• What MCP spec version does v1.2.0 target? Are prompts backward‑compatible with clients that only support tools?
• Is there test coverage for the prompts loading path? How are errors surfaced?
• What is the upgrade path from 1.1.x? Are any config changes required?

Integration Approach

Stack fit: The package targets Laravel (likely 10/11) with Nova 4. It assumes PHP 8.1+. Teams using Nova for admin panels and wanting to expose data to AI agents via MCP will find a natural fit. The new prompts feature is useful for standardising LLM interactions (e.g., “summarise this resource”, “draft an email based on this record”) without hard‑coding prompts in client applications.

Migration path: For new adopters, start with v1.2.0 directly. For existing users on 1.1.x, a minor update (composer update opscale-co/nova-mcp) should suffice; the prompts resolver is likely opt‑in via configuration. No breaking changes are indicated by the minor version bump. Review the published config for new prompt‑related keys and ensure any custom resolvers implement the expected contract.

Compatibility: Verify composer.json constraints for Laravel, Nova, and PHP. The package must be compatible with your MCP client’s supported spec version – prompts are a newer primitive, so older clients may ignore them. If you rely on prompts, confirm client support. The merge commit does not affect runtime compatibility.

Sequencing: (1) Install and configure the package in a development environment. (2) Expose a small set of read‑only Nova resources as tools and validate with an MCP client. (3) Define a few prompt templates and enable the prompts resolver. (4) Test prompt loading, error handling, and authorization. (5) Roll out to staging with representative data. (6) Production rollout with monitoring. Avoid enabling prompts for sensitive resources until authorization behaviour is confirmed.

Operational Impact

Maintenance: Low‑to‑moderate. The package is small, but MCP is evolving; expect periodic updates to align with spec changes. Prompt templates become content that must be versioned and reviewed alongside business logic. The sparse release notes mean you should monitor the GitHub repository for detailed changelogs and PR discussions.

Support: Community‑driven via GitHub issues. No commercial support is indicated. Internal ownership is needed for configuration, prompt content, and security review. Allocate a Laravel developer with Nova experience to manage upgrades.

Scaling: The MCP endpoint adds HTTP load. Tools that query Nova resources will hit the database; prompts are lightweight but may trigger tool calls. Apply rate limiting at the route level, cache prompt resolution if it involves filesystem or DB lookups, and leverage Nova’s built‑in authorization to restrict data exposure. For high‑traffic AI agents, consider a dedicated queue or read replica.

Failure modes: (1) MCP client misconfiguration – endpoint URL, auth token, or spec mismatch. (2) Prompt resolver errors – missing prompt files, invalid format, or resolver exceptions causing 500s. (3) Authorization gaps – if prompts bypass Nova policies, sensitive instructions or data could leak. (4) Performance degradation from unoptimised tool queries. (5) The merge commit’s presence suggests potential branch‑management issues, but no runtime impact is evident.

Ramp‑up: Moderate. Laravel developers can install and configure the package quickly, but understanding MCP concepts (tools, prompts, resources, client integration) requires a learning curve. Budget time for prompt design, security review, and client testing. A proof‑of‑concept with one resource and one prompt is a good starting point before broader rollout.