FEATURES

One workspace for the
full development cycle.

Understand your project, build with AI assistance, inspect performance and review every change. Work with open-source models, other providers or a local model in the same development environment.

Precise tasks. Focused execution.

Fragmentation breaks complex requests into clear tasks. Dynamic Tool Generation is our architecture direction for storing task state once and reading it on demand, helping small models work with less unrelated context. The task-tool warehouse is not yet in the current download.

Explore our approach

CODE REVIEW AND GITHUB

See every edit the assistant makes.
Then push it to GitHub.

Each file the assistant changes appears in the chat with its added and removed lines and a diff you can expand. Nothing is committed until you commit it.

Expand a file to read its diff.
See which files changed at a glance, open one for the code, then fold it away.
Publish to GitHub from the editor.
Sign in with a device code, create a public or private repository, commit and push. Progress shows in the panel and the status bar.
Check what a change may affect.
Behavior review groups saved changes into areas such as authentication, API contracts and data, and suggests checks before you stage them.

Working with GitHub Actions? Ask the assistant to create or edit a workflow file, review the diff, and push it with your code.

How source control works

workspace / ledgerILLUSTRATIVE DEMO

CHANGES YOU CAN FOLLOW

From edit to commit.

+0−0

src/accounts.ts+0−0

1-const balance = account.balance;2-return balance;3+const balance = entries.reduce(4+  (total, entry) => total + entry.amount,5+  0,6+);7+return { accountId: account.id, balance };

tests/accounts.test.ts+0−0

1+it('totals the ledger entries', () => {2+  const entries = [{ amount: 50 }, { amount: -20 }];3+  expect(getBalance(account, entries).balance).toBe(30);4+});

.github/workflows/checks.yml+0−0

1+name: Checks2+on: [push, pull_request]3+jobs:4+  test:5+    runs-on: ubuntu-latest6+    steps:7+      - uses: actions/checkout@v48+      - uses: actions/setup-node@v49+        with:10+          node-version: 2211+      - run: npm ci && npm test

Behavior review

Balance calculation changed. Check credits, debits, and an empty ledger before staging.

Preparing your changes

example / ledger · Private repository

Expand a file to inspect the diff. No repository is created.

See exactly what the AI changed

Each file the assistant edits shows its added and removed lines, with a diff you can expand and collapse in the chat.

INLINE DIFFS

Publish to GitHub from the editor

Sign in with a device code, create a public or private repository, then stage, commit and push. Progress shows while publishing.

SOURCE CONTROL

Check what a change may affect

Groups saved changes into areas such as authentication, API contracts and data, and suggests checks before you stage them.

BEHAVIOR REVIEW

Find code by symbol, text or meaning

A symbol graph built on your computer, text search and optional embeddings find the code that matters for your task.

CODEBASE INDEXING

An assistant that edits and runs commands

Ask for a change: it edits files, runs the commands you approve and reports what it did. /undo restores the files a task changed.

AI ASSISTANT

Run tests and servers in the editor

A full terminal in your workspace folder. Commands the assistant runs stream their output into the chat.

INTEGRATED TERMINAL

A native C++ analysis engine

A C++ engine parses your code for complexity analysis and the symbol graph, with 14 language grammars built in.

NATIVE CODE ANALYSIS

Big-O and N+1 query detection

Estimates each function's time and space complexity and finds N+1 queries across files. Confirm a finding with a benchmark.

PERFORMANCE ANALYSIS

Approve commands, or trust a workspace

Each command shows what will run and waits for approval, unless you allow commands for a trusted workspace. Turn that off at any time in settings.

PERMISSIONS

DEMO

A change, from search
to test run.

Follow an illustrative coding task as the assistant finds context, writes a change, and runs example checks. In your workspace, the files, commands, and results depend on your project and permissions.

Follow the setup guide

commerce / orders.jsINTERACTIVE DEMO

1export function totalsByUser(users, orders) {2  // Index once. Look up in constant time.3  const totals = new Map(4    users.map(user => [user.id, 0])5  );6 7  for (const order of orders) {8    if (totals.has(order.userId)) {9      totals.set(10        order.userId,11        totals.get(order.userId) + order.amount12      );13    }14  }15 16  return totals;17}

TERMINAL~/commerce

❯ rg -n "totalsByUser" src

src/services/orders.js:1

Lamied agent

Read the context · locating related code

Illustrative workflow · not a live benchmark

Estimates, not measurements.

Static time, space, and network findings help you investigate. They are not live latency measurements, database execution plans, or a guarantee that every change is faster.

Review what the AI writes.

AI output needs review. Use your tests, inspect changes, and measure behavior under realistic conditions before shipping.

Ready to try it on your own code?

Download Lamied

One workspace for thefull development cycle.