tidewave
Tidewave MCP dev tools for Phoenix applications. Use when setting up Tidewave, connecting AI coding assistants to a running Phoenix app, configuring MCP server access, debugging with runtime introspection tools, or troubleshooting Tidewave integration.
What this skill does
# Tidewave MCP Dev Tools
Tidewave is a dev tool by Dashbit that connects AI coding assistants to running Phoenix applications via the Model Context Protocol (MCP). It exposes runtime introspection tools: Ecto schemas, code execution, documentation lookup, log inspection, and SQL queries.
Current version: `~> 0.5` (v0.5.6 released 2026-03-13 — adds `:extra_apps` plug option)
## When connected, prefer MCP tools over web fetches
**When Tidewave MCP is connected to an Elixir/Phoenix project, route documentation, source, and runtime queries through the MCP tools — not WebFetch or hexdocs.pm.**
- `get_docs` instead of WebFetch on `hexdocs.pm/<package>/<Module>.html`
- `search_package_docs` instead of broader hex-doc searches
- `get_source_location` instead of reading `deps/<package>/lib/...` with Read
- `execute_sql_query` instead of shelling into `psql`
- `project_eval` instead of running `iex -S mix` snippets
Why: MCP tool calls return results pinned to the exact versions in the project's `mix.lock`, complete in a single round-trip, and skip HTML→markdown conversion and page chrome — significantly fewer tokens per lookup than WebFetch, and the answer is guaranteed to match the running app rather than whatever version is on hexdocs.pm today.
## Using tidewave in tier prompts
When a worker in a staged pipeline interacts with unfamiliar dependencies, the lead's prompt to that worker MUST explicitly direct the worker to verify dep APIs via tidewave MCP tools before stubbing or implementing.
Add this block to Tier 2 (test author) and Tier 3 (implementer) prompts when the work touches a dep not already exercised in the codebase:
```text
## Module docs via tidewave (PREFERRED over WebFetch or guessing)
For unfamiliar deps, verify the actual API surface via tidewave MCP before stubbing or implementing:
- `mcp__tidewave-<app>__search_package_docs` (q: "<ModuleName>") — keyword search across installed package docs
- `mcp__tidewave-<app>__get_docs` (module: "<Module>", function: "<func>") — full docstrings for a specific function
Particularly important for: <list the specific deps this tier touches>.
Tidewave hits the running BEAM's loaded modules — it returns what is ACTUALLY in the dep, not what the model remembers from training data. Guessing produces code that does not compile or behaves wrong; tidewave produces code matching the real API surface.
```
The lead lists the SPECIFIC dep names in the prompt — abstract "use tidewave" instructions are ignored. Naming the dep (`Cloak`, `Slipstream`, `Phoenix.Token`, `Mox`, etc.) and the exact MCP tool names triggers the worker to call them.
This applies whenever a tier worker touches a dependency that:
- Is new to the codebase (not already imported / used in `lib/`).
- Has changed API across a recent major version.
- The worker would otherwise stub from training-data memory.
## Installation
### New Projects (via Igniter)
```bash
mix archive.install hex igniter_new
mix igniter.install tidewave
```
### Existing Projects (Manual)
1. Add to `mix.exs`:
```elixir
defp deps do
[
{:tidewave, "~> 0.5", only: :dev}
]
end
```
2. Add plug to `lib/my_app_web/endpoint.ex` before the `code_reloading?` block:
```elixir
if Mix.env() == :dev do
plug Tidewave
end
if code_reloading? do
# ...existing code reload plugs...
end
```
3. Fetch dependencies:
```bash
mix deps.get
```
### Umbrella Projects
Apply the manual steps to the application defining the Phoenix endpoint (typically `apps/your_app_web`).
## MCP Server Configuration
Tidewave exposes an MCP server at `/tidewave/mcp` on the Phoenix app's port. The transport type is HTTP (streamable) — SSE was removed in v0.4.0. Tidewave is unauthenticated; decline any "Authenticate" prompt the editor offers.
Default URL: `http://localhost:4000/tidewave/mcp`
See `references/mcp-setup.md` for verbatim setup commands and JSON config for **Claude Code, Codex CLI, Gemini CLI, and opencode**, plus a `curl` ping for raw verification and a troubleshooting table.
## MCP Tools Reference
| Tool | Description |
|------|-------------|
| `project_eval` | Execute Elixir code within the running application runtime |
| `execute_sql_query` | Run SQL queries against the application database |
| `get_ecto_schemas` | List all Ecto schema modules with their fields and associations |
| `get_ash_resources` | List all Ash resources (when using the Ash framework) |
| `get_docs` | Retrieve documentation for modules/functions using exact project versions |
| `search_package_docs` | Query hexdocs.pm filtered to project dependencies (availability varies by framework) |
| `get_source_location` | Find module/function source code file paths and line numbers |
| `get_models` | List all application modules with their file locations |
| `get_logs` | Access server logs; supports log level filtering (added v0.5.5) |
**Note**: `search_package_docs` may not be available in all frameworks. Verify availability for your setup.
### Usage Patterns
See "When connected, prefer MCP tools over web fetches" at the top of this skill for the routing rule.
Introspect schemas before writing queries:
```
get_ecto_schemas → discover schema structure → execute_sql_query
```
Look up documentation for project dependencies:
```
get_docs for specific modules, search_package_docs for broader hex dependency searches
```
Execute and test code in the running app:
```
project_eval → run Elixir expressions against live application state
```
## Plug Configuration Options
```elixir
if Mix.env() == :dev do
plug Tidewave,
allow_remote_access: false,
inspect_opts: [pretty: true, limit: 50]
end
```
| Option | Default | Description |
|--------|---------|-------------|
| `allow_remote_access` | `false` | Allow connections from non-localhost |
| `inspect_opts` | `[]` | Options passed to `Kernel.inspect/2` for output formatting |
| `team` | `nil` | Team configuration (e.g., `[id: "my-company"]`) |
| `extra_apps` | `[]` | Additional OTP apps to include in source/module discovery (v0.5.6+) |
## CLI App (Standalone MCP Server)
The Tidewave desktop/CLI app (`tidewave_app`) runs a standalone MCP server without requiring a Phoenix application. Useful for containers, remote dev environments, or non-Phoenix Elixir projects.
- Default server: `http://localhost:9832`
- Only allows access from the same machine by default
### Installation
- **Desktop app** (macOS, Windows, Linux): https://tidewave.ai/install
- **Development CLI**: `cargo run -p tidewave-cli [-- --help]`
### When to Use CLI vs Phoenix Plug
| Scenario | Use |
|----------|-----|
| Phoenix application | `plug Tidewave` in endpoint.ex |
| Container/remote dev | CLI app or plug with `allow_remote_access: true` |
| Non-Phoenix Elixir | CLI app |
| Standalone MCP server | CLI app on port 9832 |
## LiveView Debug Annotations
Enable in `config/dev.exs` to help Tidewave map rendered HTML back to source HEEx templates:
```elixir
config :phoenix_live_view,
debug_heex_annotations: true
config :phoenix,
debug_attributes: true
```
Requires Phoenix LiveView v1.1+.
## Security
- **Dev-only**: Always guard with `Mix.env() == :dev` or `only: :dev` in deps
- **Localhost-only**: Tidewave only accepts localhost requests by default
- **No production use**: Tidewave exposes code execution and database access — never deploy to production
- **Docker/remote dev**: Set `allow_remote_access: true` only on trusted networks
- **CSP**: Tidewave injects `unsafe-eval` in `script-src` directives and disables `frame-ancestors` restrictions
## Troubleshooting
### MCP connection fails
- Verify Phoenix server is running: `mix phx.server`
- Check the port matches your configuration (default 4000)
- Try `http://127.0.0.1:4000/tidewave/mcp` if IPv6 causes issues
### Tools not appearing
- Confirm `plug Tidewave` is placed before `if code_reloading?` in `endpoint.ex`
- Restart the Phoenix server after adding the plug
- Verify the dependency is installed: `mix deps.get`
##Related in AI Agents
skill-development
IncludedComprehensive meta-skill for creating, managing, validating, auditing, and distributing Claude Code skills and slash commands (unified in v2.1.3+). Provides skill templates, creation workflows, validation patterns, audit checklists, naming conventions, YAML frontmatter guidance, progressive disclosure examples, and best practices lookup. Use when creating new skills, validating existing skills, auditing skill quality, understanding skill architecture, needing skill templates, learning about YAML frontmatter requirements, progressive disclosure patterns, tool restrictions (allowed-tools), skill composition, skill naming conventions, troubleshooting skill activation issues, creating custom slash commands, configuring command frontmatter, using command arguments ($ARGUMENTS, $1, $2), bash execution in commands, file references in commands, command namespacing, plugin commands, MCP slash commands, Skill tool configuration, or deciding between skills vs slash commands. Delegates to docs-management skill for official documentation.
reprompter
IncludedTransform messy prompts into well-structured, effective prompts — single or multi-agent. Use when: "reprompt", "reprompt this", "clean up this prompt", "structure my prompt", rough text needing XML tags and best practices, "reprompter teams", "repromptception", "run with quality", "smart run", "smart agents", multi-agent tasks, audits, parallel work, anything going to agent teams. Don't use when: simple Q&A, pure chat, immediate execution-only tasks. See "Don't Use When" section for details. Outputs: Structured XML/Markdown prompt, quality score (before/after), optional team brief + per-agent sub-prompts, agent team output files. Success criteria: Single mode quality score ≥ 7/10; Repromptception per-agent prompt quality score 8+/10; all required sections present, actionable and specific.
adaptive-compaction
IncludedAdaptive add-on policy and recovery layer that decides WHEN to compact, prune, snapshot, or fork -- replacing fixed-percent auto-compaction across Claude Code, Codex, and MCP-capable hosts. Trigger on auto-compact timing or damage: "when should I compact", "is it safe to compact now or start a fresh session", "auto-compact fires too early/mid-task", "switching to an unrelated task but the window still has space", "context rot", "answers get worse the longer the session runs", "the agent forgot the plan or my decisions after it summarized", "add a layer on top that manages context without changing the agent", raising autoCompactWindow to give the policy room, or installing/tuning a cross-tool compaction policy or PreCompact hook -- even when "compaction" is never said but the problem is context-window pressure or post-summarization memory loss. Do NOT use to summarize a conversation, build RAG, write a summarization prompt (decides WHEN not HOW), or answer max-context-length trivia.
agent-skill-creator
IncludedCreate cross-platform agent skills from workflow descriptions. Activates when users ask to create an agent, automate a repetitive workflow, create a custom skill, or need advanced agent creation. Triggers on phrases like create agent for, automate workflow, create skill for, every day I have to, daily I need to, turn process into agent, need to automate, create a cross-platform skill, validate this skill, export this skill, migrate this skill. Supports single skills, multi-agent suites, transcript processing, template-based creation, interactive configuration, cross-platform export, and spec validation.
llm-wiki
IncludedUse when building or maintaining a persistent personal knowledge base (second brain) in Obsidian where an LLM incrementally ingests sources, updates entity/concept pages, maintains cross-references, and keeps a synthesis current. Triggers include "second brain", "Obsidian wiki", "personal knowledge management", "ingest this paper/article/book", "build a research wiki", "compound knowledge", "Memex", or whenever the user wants knowledge to accumulate across sessions instead of being re-derived by RAG on every query.
skill-master
IncludedAgent Skills authoring, evaluation, and optimization. Create, edit, validate, benchmark, and improve skills following the agentskills.io specification. Use when designing SKILL.md files, structuring skill folders (references, scripts, assets), ingesting external documentation into skills, running trigger evals, benchmarking skill quality, optimizing descriptions, or performing blind A/B comparisons. Keywords: agentskills.io, SKILL.md, skill authoring, eval, benchmark, trigger optimization.