Your coding agent knows your code. Now it knows your web app too.
Connect Claude Code, Codex, or Copilot to your running app: database, logs, and live UI. Point to any element to reference it. Your agent sees what you see.
There's a better way to work on web apps with AI
To get the most from your coding agent, you need to give it the right tools.
The question is how.
Multiple tools, multiple configs
Hunt down separate tools for browser, docs, database. Configure each. Debug when they clash.
Generic tools
Playwright, Context7, and others work, but they can't see inside your app. So your agent has to fill in the gaps.
Text only
Describe UI elements to the agent: "The button in the header—no, the other header...". Copy and paste stacktraces.
One install, one config
Browser, docs, database, logs, code execution, all in one package. Set up in minutes.
Framework-integrated tools
Trace pages and components to source. Get docs for your exact dependency versions. Run code in your running app.
Text + point and click
Include UI elements in your prompts with a click. Fix error pages by pressing a button. Do more, type less.
Features
A full-stack coding agent, in the browser
Tidewave runs your coding agent in the browser. The agent works across your entire stack, from UI to the database, and you see every change as it happens. No more translating between what you're looking at and what the agent knows.
Point and click prompting
Want to change something? Click it. Tidewave traces that element back to the source code and hands all relevant UI metadata to the agent, so you don't have to. It works across frontend and backend tools.
Verify changes in your browser
You and the agent share the same browser window. When the agent makes a change, it navigates your app to verify it works. When something breaks, it sees the same error page you do and can offer to fix it.
Deep integration with your web framework
Give your coding agent new tools to access everything your app knows.
Get better results, less back-and-forth.
We call this Runtime Intelligence.
Query your database
Uses your app's existing connection. Direct access to your models, no need to grep for them.
Read your logs
When something is not right, Tidewave sees the same stacktrace you do, automatically. No more copy-paste.
Execute code in your running app
Tests methods, checks values, explores APIs. Like having a REPL for agents.
Access documentation
Pulls docs for your exact library versions, not whatever a web search returns.
Works with the agents you already use
You can also bring your own API key (Anthropic, OpenAI, OpenRouter) or any ACP-compatible coding agent.
Tidewave enhances Claude Code, making it 2x more accurate and 45% faster on average for full-stack changes
Read our report →Pricing
Pro and Teams subscriptions unlock all Tidewave features. You'll also need one of: Claude Code, OpenAI Codex, GitHub Copilot, or bring your own API key (Anthropic, OpenAI, OpenRouter).
Tidewave MCP
Database access
Logs and runtime introspection
Code execution
Documentation access
Tidewave Web
In-browser coding agents
Point and click prompting
Full-stack web framework integration
Web accessibility diagnostics
Figma Dev Mode support
Plus everything in OSS
Team features
Enforce settings org-wide
Centralized team billing
Early adopter pricing
Coming soon
Single Sign On (SSO)
Usage analytics and leaderboards
Plus everything in Pro
Price excludes taxes. Pro's free trial has a limit of 20 user messages per month.
Get started in 3 steps
Download Tidewave
Get the app and sign in with GitHub. Free trial, no credit card.
Add the package to your app
Works with Rails, Phoenix, Django, React, TanStack, Next.js, and more.
Connect your agent
Open your app in Tidewave and select Claude Code, Codex, or Copilot.
Tidewave differs from traditional coding agents by operating directly
within your browser, maintaining shared context between you, the agent,
and your app. Conventional coding agents require you to constantly describe
UI states, copy error messages, and switch between tools, while Tidewave
can see your actual web pages, automatically map UI elements to their
corresponding code, query your database, and test changes in real-time.
This eliminates the tedious back-and-forth translation process and allows
the agent to build and validate complete features without losing context,
making development far more seamless and efficient.
We believe the future of coding agents will be domain-specific and
deeply integrated to your stack, rather than generic horizontal solutions.
Tidewave offers the following advantages compared to generic browser tools:
- Made for your stack: Tidewave integrates with front-end and backend frameworks to map DOM elements to source code, reducing the amount of guess work done by agents.
- Inspector-enabled: Tidewave's inspector allows you to interact with page elements for point-and-click prompting.
- Error handling: Tidewave automatically recognizes error pages and feeds stacktraces, logs, framework metadata to the agent.
- Cross-browser support: Tidewave runs in your favorite browser, instead of a select few, allowing you to reuse your sessions and reduce context-switching.
We have support for Claude Code, OpenAI Codex, GitHub Copilot subscriptions, as well as any coding agent that implements ACP. You may also bring your own API key for Anthropic, OpenAI, and OpenRouter.
We log basic request metadata (timestamps, model used, token counts).
Prompts and messages are not logged unless you explicitly opt-in.
We don't store tool results. Note the underlying model provider you
use may store data separately depending on your user agreement with them.
See our Terms of Service and
Privacy Policy.
Yes! The Pro plan includes a free trial, no credit card required, with a limit of 20 user messages per month.
We are working on enterprise plans and we would love your feedback on how to best shape them according to your needs. Please send us an email with your needs.
Please read our
security guidelines, which includes common concerns and isolation techniques
such as Docker support.
Note Tidewave is a development tool and should not run in production.
Some components of Tidewave are open source. In particular, we have open sourced the "Runtime Intelligence" component for each supported language/framework, which you can use directly from most AI editors and third party coding agents through the Model Context Protocol (MCP), without a Tidewave subscription. Check out our Organization on GitHub for all supported languages and frameworks.