Fire in da houseTop Tip:Paying $100+ per month for Perplexity, MidJourney, Runway, ChatGPT and other tools is crazy - get all your AI tools in one site starting at $15 per month with Galaxy AI Fire in da houseCheck it out free

tidewave_phoenix

MCP.Pizza Chef: tidewave-ai

While you're building a Phoenix web application, this tool lets your coding assistant run real code inside your running app, check the actual logs, look up documentation for the exact library versions you're using, and query your development database directly. It works inside editors like Cursor and VS Code, needs no separate account or key, and installs by adding one small package to your project. It comes from the team behind Tidewave, a development environment built for Phoenix and Rails applications.

Coding
Data

Use This MCP server To

Run real code inside my running app to test an idea Check my app's live logs to see what broke Look up documentation for the exact library versions I'm using Query my development database right from my coding assistant Jump straight to the file and line where code lives List every database model in my project at once

README

Tidewave Phoenix

Tidewave Phoenix is an MCP server that provides runtime-level tools for developing Phoenix apps using coding agents.

Your agent will be able to use this MCP server to talk to your running Phoenix app in development to:

  • execute code in the context of the running app (like an IEx session for agents)
  • read the app's live logs
  • query your development database
  • get source locations of modules and functions
  • read documentation pinned to the exact hex package versions your project depends on

This MCP server is an open-source component of Tidewave, the agentic development environment for Phoenix and Rails.

You can use this project as a standalone MCP server or integrated with the Tidewave product by following the installation instructions below.

Installation

1. Add the Tidewave hex package to your app

Option 1: Manually

Add the tidewave package to your mix.exs:

def deps do
  [
    {:tidewave, "~> 0.6", only: :dev},
    {:phoenix, ...},
  ]
end

Then, for Phoenix applications, go to your lib/my_app_web/endpoint.ex and right above the if code_reloading? do block, add:

+  if Mix.env() == :dev do
+    plug Tidewave
+  end

   if code_reloading? do

Tip

Tidewave works best with Phoenix LiveView v1.1 or later. Once you update it, make sure to enable the following options in your config/dev.exs:

config :phoenix_live_view,
  debug_heex_annotations: true,
  debug_attributes: true

Those are enabled by default for Phoenix v1.8+ apps.

Option 2: Using Igniter

Alternatively, you can use igniter to automatically install Tidewave MCP into an existing Phoenix application:

# install igniter_new if you haven't already
mix archive.install hex igniter_new

# install tidewave
mix igniter.install tidewave
Umbrella projects

For umbrella projects, you can follow the manual steps above in the application that defines your Phoenix endpoint (typically apps/your_app_web).

In non-Phoenix applications

Tidewave can be used as a MCP in any Elixir project. For example, you can use bandit (and tidewave) in dev mode in your mix.exs:

{:tidewave, "~> 0.6", only: :dev},
{:bandit, "~> 1.0", only: :dev},

And then adding an alias in your mix.exs:

aliases: [
  tidewave:
    "run --no-halt -e 'Agent.start(fn -> Bandit.start_link(plug: Tidewave, port: 4000) end)'"
]

Now run mix tidewave

2. Add the Tidewave MCP to your agent/editor

Add the Tidewave MCP server to your editor or MCP client configuration as the type "http" (streamable), pointing to the /tidewave/mcp path and port your web application is running at. For example, http://localhost:4000/tidewave/mcp.

We also have specific instructions for:

Usage

As with any other MCP server, your agent will call the tools exposed by the Tidewave MCP whenever it sees fit. But you can also prompt it to call them explicitly.

Available MCP tools

project_eval

Evaluates Elixir code within your running application, giving the agent access to your runtime, dependencies, and in-memory data. It's like an IEx for the agent.

project_eval demo

Your agent can use it when it would rather run code than assume behavior, grounding its next step in what the running app actually does. For example, calling a function to see what comes back or reproducing a failing code path against live app state to debug it.

execute_sql_query

Executes a SQL query within your app's development database.

execute_sql_query demo

Your agent can use it to run any SQL against your development database. Useful for the agent to verify the result of an action.

get_docs

Get the documentation for a given module/function. It consults the exact versions locked in your project's mix.lock, ensuring you get correct information.

get_docs demo

get_logs

Reads logs written by the server.

get_logs demo

Your agent can use it to see what happened after a request. For example, reading the request log and backtrace when something misbehaves.

get_source_location

Get the source location for a given module/function, across both your app and its dependencies.

get_source_location demo

Your agent can use it to jump straight to where a module/function is defined, by file and line, instead of grepping for it, including when the definition lives in a hex dependency.

get_ecto_schemas

Lists all Ecto schema modules and their file paths.

get_ecto_schemas demo

get_ash_resources

Returns all Ash domains and their resources for the current project.

get_ash_resources demo

Only available if you are using Ash.

Troubleshooting

Using multiple hosts/subdomains

If you are using multiple hosts/subdomains during development, you must use *.localhost, as such domains are considered secure by browsers. Additionally, add the following immediately @session_options definition in your lib/your_app_web/endpoint.ex:

@session_options [
  # ... your configuration
]

if code_reloading? do
  @session_options Keyword.merge(@session_options, same_site: "None", secure: true)
end

The above will allow your application to run embedded within Tidewave across multiple subdomains, as long as it is using a secure context (such as admin.localhost, www.foobar.localhost, etc).

Content security policy

If you have enabled Content-Security-Policy, Tidewave will automatically enable "unsafe-eval" under script-src in order for contextual browser testing to work correctly. It also disables the frame-ancestors directive. This is done only in the environments that Tidewave is loadead (development by default).

Configuration

You may configure the Tidewave plug using the following syntax:

  plug Tidewave, options

The following options are available:

  • :allow_remote_access - Tidewave only allows requests from localhost by default, even if your server listens on other interfaces, for security purposes. Read our security guidelines for more information and when to allow remote access (if you know what you are doing)

  • :allowed_origins - a list of values matched against the Origin header to prevent cross origin and DNS rebinding attacks. Each value must be a string of shape [scheme:]//host[:port], where both scheme and port are optional. The host may also start with "*". Example: ["//localhost:8000", "//*.test"]

  • :inspect_opts - custom options passed to Kernel.inspect/2 when formatting some tool results. Defaults to: [charlists: :as_lists, limit: 50, pretty: true]

  • :team - set your Tidewave Team configuration, such as team: [id: "my-company"]

  • :toolbar - controls whether the Tidewave toolbar is injected into your HTML pages. Defaults to true

  • tmp_dir - temporary directory Tidewave uses for screenshots and recordings. It must be a relative directory to the current application root. Defaults to tmp, storing files under tmp/tidewave/screenshots and tmp/tidewave/recordings

License

Copyright (c) 2025 Dashbit

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

tidewave_phoenix FAQ

Can I use this to check what went wrong after an error in my app?
Yes — it can read your app's live logs so your assistant sees exactly what happened, including the backtrace, right after the error occurs.
Can I use this to look up documentation for my project's dependencies?
Yes — it pulls documentation matched to the exact versions locked in your project, so you get accurate answers instead of generic ones that might not fit your setup.
Which apps does this work with?
It works with any assistant that supports MCP, including Cursor, VS Code, Claude Code, Codex, and Neovim.
Do I need an account or a paid key to use it?
No — it runs alongside your own project during development, with no external account, sign-up, or subscription required.
Is this easy to set up if I'm not very technical?
Not really — it's built for people already building an Elixir and Phoenix application, and setup means adding a package to your project and a couple of lines of code, which is a developer task.
Will this work on my live, public website?
No — it's meant to run only in development, connected to the copy of your app running on your own computer, not on a public production site.
Can I use this with Ruby on Rails instead of Phoenix?
This particular version is built for Phoenix and Elixir projects; the same company also makes a separate version for Rails.
Does it change my database or app without me asking first?
It only runs code or database queries when your assistant is prompted to, so keep an eye on what it's about to do if you're touching anything you can't easily undo.