// mcp for laravel

LaraBench MCP for your own Laravel apps, and for public Laravel search.

Use the desktop MCP to give Claude, Codex, Cursor, or another client controlled access to your own Laravel connections, logs, saved items, and run history. Use the public MCP — no account, no install — to search Laravel jobs from the last 30 days, plus ecosystem news and package releases from the last 7 days.

Add the desktop MCP Add the public MCP
Desktop MCP setup in LaraBench: client buttons, tools, and per-connection AI access in one screen.

// why there are two

Two servers, because they answer different questions.

The desktop server is the reason LaraBench exists: it is how an AI client reaches a real Laravel app without you handing over the whole machine. The public server is a much smaller thing — LaraBench already indexes Laravel jobs and ecosystem news for the website, so exposing that index over MCP costs us nothing and saves you a browser tab. It is read-only, it touches none of your data, and you never need the desktop app to use it. If you only came here for one of them, it is almost certainly the first.

// desktop mcp

Connect AI clients to your Laravel apps.

The desktop server exposes only the LaraBench connections and local data you select. Premium adds run and save tools, with a separate access level for every connection.

Download LaraBench

// public mcp

Search recent Laravel jobs, news, and releases.

The public server needs no account or desktop app. It is read-only and returns LaraBench detail pages: jobs cover the last 30 days, news and releases the last 7 days. Requests are capped at 60 per minute per IP.

https://larabench.com/mcp

// choose the right server

Public MCP or desktop MCP?

Best for

// public Recent Laravel jobs, articles, and package releases.

// desktop Your Laravel connections, logs, saved items, and run history.

Install

// public No desktop app. Add the remote URL in a compatible MCP client.

// desktop Install LaraBench and connect your AI client from the app.

Data

// public Public LaraBench jobs and news only.

// desktop Your exposed connections, saved library, history, logs, jobs, news, and profile.

Writes

// public None. Read-only.

// desktop Premium tools can run snippets, run Artisan, run saved items, and save snippets.

URLs

// public A larabench.com detail page plus the original source URL.

// desktop Local app context plus LaraBench jobs/news URLs.

// public tools

Tools available on the public MCP.

These tools return compact summaries, a LaraBench detail URL, and the original source URL — the apply link for a job, the article for a news item, the release notes for a release. They never return saved snippets, logs, or anything from a desktop app.

search_laravel_jobs

Search recent Laravel jobs indexed by LaraBench, with full description, salary, seniority, and employment type. Results are capped to the last 30 days and carry both a LaraBench job page and the original apply URL.

querytechnologiesworkArrangementlocationscountriessalaryMinimumsalaryCurrencyseniorityemploymentTypeslimit

search_laravel_news

Search recent Laravel ecosystem articles and package releases. Results are capped to the last 7 days and carry both a LaraBench news page and the original source URL.

querytypestopicsprojectspackagessourceschannellimit

// desktop tools

Tools available through the desktop app.

The desktop MCP reads only the LaraBench data and connections you expose. Run and save tools require Premium, the open app, and permission on the selected connection.

Read and search tools

list_connections

List the Laravel connections exposed to AI tools and show each connection's access level.

list_saved_items

Browse saved Tinker snippets and Artisan commands by folder, tag, search, kind, or pinned state.

get_run_history

Read recent LaraBench runs with code, command, output, exit code, duration, and query count.

read_logs

List Laravel log files or read entries from a selected file. SSH and Docker logs need the app open.

search_laravel_jobs

Search LaraBench jobs from the desktop MCP. Free searches cover 30 days; signed-in Premium extends to 90 days.

search_laravel_news

Search LaraBench news and package releases from the desktop MCP. Premium extends the search window to 90 days.

get_developer_profile

Return your saved developer profile and the current job filter options when you are signed in.

Run and save tools, Premium

run_snippet

Run a Tinker snippet against a connection and return output, exit code, duration, and query analysis.

run_artisan

Run an Artisan command against a connection. Artisan runs need Full access because they cannot be dry-run safely.

run_saved_item

Run a saved snippet or Artisan command, filling any {placeholder} values first.

save_snippet

Save a Tinker snippet or Artisan command back to your LaraBench library.

// add to your agent

Add LaraBench MCP to your client.

Cursor can open the public server in an install prompt. The Claude and Codex buttons copy their official CLI commands. The desktop app writes its local stdio config for each supported client.

Public MCP

Add to Cursor

No API key or account is required.

  1. 1Add a remote HTTP server named larabench-public.
  2. 2Use https://larabench.com/mcp as the URL.
  3. 3Leave authentication empty. The server is public and read-only.
  4. 4Ask for recent Laravel jobs, news, or package releases.
{
  "mcpServers": {
    "larabench-public": {
      "url": "https://larabench.com/mcp"
    }
  }
}

Desktop MCP

  1. 1Install and open LaraBench.
  2. 2Create the local, Docker, or SSH connections you want to expose.
  3. 3Open the MCP Server page inside the app.
  4. 4Set each connection to Off, Read only, Dry run, or Full.
  5. 5Click the button for Claude Code, Claude Desktop, Codex, or Cursor, or use Show config for another MCP client.
{
  "mcpServers": {
    "larabench": {
      "command": "path shown by LaraBench",
      "args": ["--mcp-stdio"]
    }
  }
}

// safety

Read-only public search. Per-connection controls on desktop.

Public MCP results are anonymous and read-only. They are limited to recent public LaraBench jobs, news, and release pages.

Desktop connections start from explicit AI access levels: Off, Read only, Dry run, or Full. Production starts hidden.

Dry run cannot be loosened by a tool call. A write against Full production still waits for a human confirmation in LaraBench.

AI-driven desktop runs are recorded in History with the same output, exit code, timing, and query context as your own runs.

Where credentials live, what the local server binds to, and how its token is scoped