AI-native IDE

Your AI writes
the code.

You review
and ship.

Anide is built around the AI workflow — terminal for Claude Code or Codex, diff view to review every change, and full MCP context so your AI always knows the full picture.

Join Waitlist Star on GitHub

How it works

Intent. Execute. Review. Ship.

Anide wraps AI agents in a review-first workflow. You direct. AI executes. Every change surfaces as a diff. You commit what's right.

Describe intent

Tell the AI what you want built. It reads your codebase, schema, and API responses — the full picture, not just one file.

AI executes

Claude Code or Codex runs in the integrated terminal. It writes files, runs commands, and calls tools.

Diff appears

Every change surfaces immediately as a Git diff. You see exactly what the AI touched — line by line.

You review

Accept, reject, or adjust. You are the senior engineer on the PR — the AI is the contributor.

Commit and ship

Stage what you want, write the commit message, push. The loop resets for the next session.

MCP context

Your AI sees the whole stack.

Most AI coding tools only see the open file. Anide gives your agent MCP access to your APIs, databases, cache, storage, Git history, and running containers — so it has full context before it writes a single line.

These aren't separate tools for you to manually operate. They're context providers for your AI. The agent can look up the live schema, test the endpoint it just built, and check what's cached — all in one session.

API Client

Test the endpoint the AI just built. Response body, headers, and status all feed back into context.

Database

Read the live schema, run queries, inspect what's in the tables before writing a single migration.

Cache

Browse Redis keys and TTLs. AI knows what's cached before it touches the invalidation logic.

Object Storage

List S3 buckets and files. Upload paths, presigned URLs, and bucket policies are all visible.

Git

Full commit history, diffs, and branch state. AI sees what changed and what is pending review.

Docker

Container status, logs, and Compose config. AI knows whether your services are actually running.

Review first

Every change is a diff you approve.

Built for engineers who want AI speed without AI blind spots. Every file the agent touches surfaces as a Git diff. You review, you decide, you ship.

diff viewer

+4 −1

src/services/user.ts · +4 −1

- async function createUser(email: string) {

+ async function createUser(email: string, name: string) {

const user = await db.users.create({

+ name,

email,

createdAt: new Date(),

});

+ await notifySlack(`New user: ${name} <${email}>`);

return user;

}

Unified & split view

Toggle between a unified diff and side-by-side comparison. Large diffs paginate automatically.

Stage what you want

Accept the whole file or cherry-pick hunks. Commit exactly the changes you have reviewed and approved.

Nothing auto-commits

Changes live in the working tree until you stage and commit. You decide when the AI's work ships.

Local first

Your keys. Your machine. Your code.

Anide runs entirely on your machine. Your API key goes straight to the model — no cloud proxy, no data collection, no account required. Your codebase never leaves your disk.

Bring your own API key — Anthropic, OpenAI, or any compatible model
No telemetry. No cloud storage. No mandatory account.
Open source and auditable
Works offline for everything except model inference calls

Rust backend, native OS WebView, no bundled Chromium. Starts in under a second. Ships as a tiny binary. Dramatically less RAM than any Electron alternative.

TauriRustSvelteKitCodeMirror 6gitoxidebollard

< 1s

startup

~10 MB

binary size

none

Chromium bundled

Human + AI

Designed for the human in the loop.

Anide isn't just a place for AI to run. It's where you browse docs, inspect API responses, query the database, and read container logs — everything a senior engineer does while reviewing an AI's work before it ships.

Browse docs

Every markdown file in the repo — agent instructions, ADRs, changelogs — readable and editable without leaving the app.

Review diffs

Unified and split diff view. Stage individual files or specific hunks. Nothing ships until you've approved it.

Run the terminal

Integrated terminal for Claude Code, Codex, or any CLI. Watch the AI work, interrupt when needed, run commands yourself.

Check containers

Stream logs, restart services, run compose up/down. Diagnose what the AI broke before writing a bug report.

No more Electron bloat

One app. Six bloated tools gone.

Every tool in your current stack is a separate Electron app — each bundling its own Chromium, eating 300 MB of disk and hundreds of MB of RAM before you've done anything. Anide replaces all of them with one native desktop app built on Tauri.

Postman API Client

Cloud-first, account-gated, credentials stored on their servers. Anide is local-first with .env-aware requests.

DBeaver / Compass Database

Java app with a dated UI. No integration with the rest of your workflow.

Redis Insight Cache

Redis-only, Electron. One more app in the dock for a single use case.

MinIO Browser Object Storage

Browser-based, no API or container integration. Another disconnected silo.

GitHub Desktop Git

Git only. Anide has Git as part of a connected, AI-aware developer workflow.

Docker Desktop Docker

Bloated, controversial licensing. Anide is a lightweight companion, not a daemon replacement.

apps replaced

Postman, DBeaver, Redis Insight, MinIO, GitHub Desktop, Docker Desktop

~2 GB

RAM saved

Six Electron instances vs one native Tauri app using the OS WebView

Chromium copies bundled

Tauri uses the OS WebView — no bundled browser engine

The AI writes. You ship.

Anide is in early access. Join the waitlist to be notified when builds are available.