
Tidewave Phoenix
io.github.tidewave-aiPublished by GitHub account tidewave-aiv0.8.0Updated Oct 24, 2026
Runtime-level MCP server for Phoenix development: run code, read logs, query the DB, read docs.
Overview
Runtime-level MCP server for Phoenix apps that lets a coding agent evaluate Elixir code, read live logs, query the dev database, and look up docs and source…
- What you can do
- Evaluate Elixir code inside the running Phoenix app, like an IEx session for the agent
- Read the app's live server logs, including request logs and backtraces
- Run SQL queries against the app's development database
- Look up documentation pinned to the exact hex package versions in mix.lock
- Find the file and line where a module or function is defined, including in dependencies
- When to use it
- Useful when a coding agent works on a Phoenix application and should verify behavior against the running app instead of guessing: reproducing failing code paths, checking what a function returns, inspecting logs after a request, or confirming data in the development database.
- Requirements
- The tidewave hex package added to the app as a dev-only dependency, plus a Tidewave plug in the Phoenix endpoint. The app must be running in development; the MCP client connects over streamable HTTP to the app's /tidewave/mcp path and port (for example localhost:4000), or via the mix tidewave.proxy STDIO command when ports vary, such as with worktrees. Phoenix LiveView 1.1 or later is recommended.
Installation
In SourceWeft
- Open Tidewave Phoenix in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
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:
Then, for Phoenix applications, go to your lib/my_app_web/endpoint.ex and right above the if code_reloading? do block, add:
[!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: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:
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).
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:
[!TIP] If you are using worktrees, you are likely running your web server on different ports, and therefore there isn't a single host and port combo you can use.
In such cases, you may want to add
mix tidewave.proxyas STDIO MCP instead, which adds a "port" parameter to all tool definitions, and is responsible to dispatch to the correct application.
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.
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.
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_logs
Reads logs written by the server.
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.
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.
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:
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:
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. Use this option with extreme care and, once enabled, ensure you have an authentication proxy in front of your app. -
:allowed_origins- a list of values matched against theOriginandHostheaders 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 "*". Only the host is compared against theHostheader. IP addresses,localhost, and*.localhostare always allowed as hosts. When not set, defaults to the host of your Phoenix endpoint URL. Example:["//localhost:8000", "//*.test"] -
:inspect_opts- custom options passed toKernel.inspect/2when formatting some tool results. Defaults to:[charlists: :as_lists, limit: 50, pretty: true] -
:team- set your Tidewave Team configuration, such asteam: [id: "my-company"] -
:toolbar- controls whether the Tidewave toolbar is injected into your HTML pages. Defaults totrue -
tmp_dir- temporary directory Tidewave uses for screenshots and recordings. It must be a relative directory to the current application root. Defaults totmp, storing files undertmp/tidewave/screenshotsandtmp/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.
Source: README.md at commit 64056ef
Tools
0Version history
1- v0.8.0LatestSep 16, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts