Setting 01
Supervised
Every action waits for you to check it first.
- Reading
- Asks first
- Editing
- Asks first
- Commands
- Asks first
Nx IDE · for software teams
Meet your AI team.
Your developers explain a task in everyday words. Nx IDE's AI helpers read the project, agree a plan, write the code and test it, right inside VS Code, the code editor most developers already know. Use the AI model you prefer, even one running on your own computers. By default, nothing changes until a person says yes.
31 AI specialists built in Works with 26 AI providers Nothing sent to Nexgile
It arrives as a VS Code add-on named Nexgile Code, and a Nexgile license key turns it on.
Part of one platform
Nx·Studio builds your digital workers. Nx IDE helps your developers code with AI. The Cognitive Engine handles routine decisions it has proven it can get right.
How the three products work together In one sentence
AI helpers inside the code editor.
Your developer explains the task the way they'd tell a colleague. The helpers read the project, plan, write the code and run the tests, and by default they ask before changing anything.
Why teams use it
Speed, without skipping the checks.
With many AI coding tools, people still check every line themselves. In Nx IDE's team mode (Orchestrator), one helper writes the change and two more test and review it, and they're built to ask before anything goes out.
Made for
Teams that answer for every change.
Developers in banks, hospitals, government and other careful organizations, plus teams updating older systems. They get AI help with sign-offs built in, controls an auditor can inspect, and freedom to switch AI companies.
One task, start to finish
What happens when you hand it a task?
Five steps. At each one, you choose how much it may do without asking.
01
01
Tell it the task
Write the request in everyday English. If you like, pick a specialist, such as Plan to map out the work or Debug to track down a problem. Then point it at the files that matter, PDFs and Word or Excel files included.
02
02
Settle on a plan
On bigger jobs, the Plan specialist drafts a plan first. Comment on any part and just that part gets rewritten. Building waits until the plan is approved, by you or, on a more hands-off setting, after a countdown.
03
03
Say yes, or let it carry on
Out of the box, every change and command waits for a person's OK. A five-step setting called the Delegation Dial, plus rules you write, can let some actions go ahead alone. Before code is saved, shared or released (committed, pushed or deployed), the AI is built to pause and ask. A deny rule stops an action at every level.
04
04
Undo at any point
It keeps a save point of your files at every step (a checkpoint), so one click takes you back to any earlier moment. A running meter shows the estimated cost along the way.
05
05
It checks its own work
By default, after editing files it reviews the work, builds the project and runs your tests, then shows exactly what changed.
your choice of AI
Choose your AI. Change your mind.
Pick from 26 AI providers, from the big AI companies to services that host open models, or run a model on your own computer. Switch whenever it suits you, so you're never tied to one supplier.
Run it on your own machines
Tools such as Ollama or LM Studio let you run AI models on your own computers, so your code never leaves your hardware.
Reuse your existing AI plan
Already pay for an AI subscription? Use it here, or use the AI models that come with VS Code.
A different AI for each role
Give each specialist its own AI model, so planning, coding and reviewing each use the one that suits them best.
Keys kept out of sight
The access keys for your AI accounts live in your computer's secure storage, never in a settings file.
privacy
Your code doesn't come to us.
Code goes straight from your editor to the AI provider you chose. Nothing of ours sits in the middle, Nx IDE sends us no usage data, and even the license check works without the internet.
Nothing reported back
Nx IDE sends us no usage reports (what the industry calls telemetry). We never see your code, what you type or how your team uses it.
A direct line
Each request goes straight to the AI provider you picked, or stays on your computer when the model runs there.
A version for US government work
A separate TAA edition, made for the US Trade Agreements Act, leaves out restricted AI providers and models altogether, and they can't be turned back on.
Keep files off-limits
Name files or folders in a special “ignore” list (.nexgileignore) and the AI won't open them.
knows your code
It learns your code before touching it.
A feature called Cortex draws a private map of your code on your own computer. It shows how the pieces fit and what a change might break, before anyone makes it, and it reads older languages such as COBOL too.
See the ripple effects
Find out what a change would affect before you make it. The map is built on your own computer, with no internet connection.
Old systems included
COBOL and modern languages share one map, so older systems aren't left out.
A memory for your project
It keeps notes between sessions and can look up what was decided in earlier chats.
Reads your documents
It reads PDFs, Word and Excel files and web pages, and with your permission it can look into other code projects too.
guardrails
Limits the software itself enforces.
On top of sign-offs, Nx IDE rejects unsafe file changes and can fence off where the AI may write. Turn on the optional sandbox and your own operating system blocks its commands from touching anything outside your project.
A save point every step
Switched on from the start. The AI keeps a checkpoint of your files as it goes, so one click rolls you back.
Fenced-off folders
Rules can restrict where the AI makes changes, and automatic tasks can be locked to chosen folders. The software holds that line.
A walled-in sandbox
An option on macOS and Linux. Commands can change files only in your project and folders you approve, and can be kept off the network.
No half-finished files
Incomplete files never get saved. A change that was cut short, or that skips code with a note like “rest of the file unchanged”, is turned down.
specialists
A specialist for every kind of task.
31 specialists come built in, from Plan and Code to Tester, Reviewer and Shipper, each with one part of the job. Create your own, or add skills and tool connections from the built-in marketplaces.
Ten core roles, 21 experts
The core roles cover the main steps, and 21 ready-made experts handle everyday jobs, from tidying code to reshaping a database.
Design your own
Write a new specialist down in a short text file, and your team reviews and tracks it like any other change.
One-click skills and tools
Add skills (reusable instructions) and tool connections from the built-in marketplaces, or write one yourself and reuse it.
Learns your house rules
Instruction files kept with your project, such as AGENTS.md, teach every specialist how your team likes things done.
Orchestrator mode
You lead. The AI team does the legwork.
In Orchestrator mode, one helper (or a small group) writes the change, two others check it separately, and a person decides if it goes out.
01 Code, or a team
Write
One helper, or 2 to 5 working together
02 Tester · Reviewer
Check
At the same time, each on its own
03 You
Your OK
You're asked: ship it?
04 Shipper
Send
Saved to its own copy of the code (a branch)
Fresh eyes on every change
The tester and reviewer start from scratch, with just the request and the changed files, so the writer's thinking can't sway them. Neither one can edit files.
Two more tries, then you
When tests fail or fixes are needed, a new specialist tries again, no more than twice. After that, the decision is yours.
One button stops everyone
A single Stop button halts every AI helper in the chain at once, and a map shows who's doing what.
Bigger jobs
Agent Manager
Run several AI chats on one screen and keep an eye on each helper's progress, cost and questions.
Bigger jobs
Agent Teams
On bigger jobs, 2 to 5 AI helpers (agents) divide the work, say the server side and the screens people use. They message each other and share one to-do list while you watch.
Bigger jobs
/batch
Split a big change into as many as 30 small tasks that run side by side, each in its own copy of the code. Each can become a draft pull request (a proposed change) on GitHub if you like.
Bigger jobs
/review
As many as five AI reviewers go over your changes together without editing a thing. Fixes begin only when you choose them.
The Shipper is told to save work to its own branch and never overwrite anyone else's. “Ship?” is a question for you: at the Supervised and Trust Reads levels it always waits for an answer. At higher levels, add a deny rule if shipping should always wait for a person.
Controls
Your review process, built in.
The same sign-offs your team expects for any code change, applied to the AI helpers too.
01
Sign-off
Sign-off by a person is switched on from the start.
02
Your rules
Write rules that allow an action, ask first or block it. Every allowed action shows which rule allowed it.
03
Pauses before it ships
Before code is saved, shared or released, a database changes or anything is deleted, the AI is built to pause and ask. A deny rule turns that into a hard stop at every level.
04
Checks you write
Scripts you write (hooks) can run at set moments, for example so your quality checks must pass before a task counts as finished.
05
Locked settings
No AI helper can change your permission, hook or add-on (plugin) files, and the main helper asks before touching its own rule files.
06
Split duties
The helpers that test and review code have no power to edit files.
Tool access, rules, hooks and the sandbox are enforced by the software itself, not just requests the AI is asked to respect.
Questions from security
Three questions your security team will ask.
Who said yes to that change? What did it cost? Who could touch the code? Nx IDE has a clear answer to all three.
Q1
Who said yes to this change?
A person, or a setting or rule that a person picked. Sign-off is on from the start, and the AI is built to pause and ask before code is saved, shared or released.
Q2
How much did it cost?
A running meter shows each task's estimated cost, and the Usage & Cost dashboard splits spending by day, AI model and project, with budget alerts. Optional limits pause the work until you decide.
Q3
Who had access to the code?
Only what your settings and rules permitted. Deny rules cover every AI helper, even ones started by another helper. None of them can change your permission or hook files, and the tester and reviewer can't edit files at all.
Most safeguards are plain settings or rule files. Your security team can read them, keep them with your code and map each one to the controls they already run.
Visit the Trust Center how much it does alone
Give it more freedom as trust grows.
A single dial below the chat box decides how much the AI may do without asking. Start with everything checked, then turn it up when you're ready.
Worth knowing: save points (checkpoints) start switched on at every level. Above Supervised, helpers started by another helper (such as the Orchestrator's specialists) get every automatic approval. Your deny rules still apply to them.
More checks More freedom
Setting 02
Trust Reads
It reads files in your project without asking. Edits and commands still need your OK.
- Reading
- Automatic
- Editing
- Asks first
- Commands
- Asks first
Setting 03
Trust Edits
It reads and edits files without asking. Routine questions, plan approval included, take the AI's first suggestion after a 60-second countdown. Commands still need your OK.
- Reading
- Automatic
- Editing
- Automatic
- Commands
- Asks first
Setting 04
Trust Execution
Reads, edits and commands go ahead alone, as do tools you've marked as always allowed. Starting more helpers, editing protected files or files outside the project, and a few other actions still ask. Deny rules still block.
- Reading
- Automatic
- Editing
- Automatic
- Commands
- Automatic
Setting 05
Autonomous
Everything goes ahead without asking, with three exceptions. Your deny rules still block. Edits to protected files or files outside the project still ask, and so do a few actions such as fetching web pages.
- Reading
- Automatic
- Editing
- Automatic
- Commands
- Automatic
Who's on the team
Ten roles, one team.
Each core role owns one part of the job, and the Orchestrator can bring several of them together on a single task.
Plan
Maps out the work before any code is written.
Code
Writes new code and improves what's there.
Chat
Explains your code and answers questions.
Debug
Tracks down bugs and fixes them.
Orchestrator
Runs the team: writer, tester, reviewer, shipper.
Tester
Tests each change on its own terms.
Reviewer
Reviews the code with fresh eyes.
Shipper
Saves and sends off approved work.
Spec Bootstrap
Asks you questions, then writes a project guide (wiki).
Spec Sync
Updates the guide whenever the code moves on.
And 21 ready-made experts for everyday jobs
RefactorTest GenDoc GenSecurityAPI ModDep UpgradeIssue FixerPR FixerMerge ResolverDocs ExtractorIssue InvestigatorIssue WriterDB MigratorDB ArchitectCode ArchitectDevOpsAPI TesterCode ReviewerTranslateUI DesignerDev
Need another? Describe it in a short text file (Markdown) and it joins the list, for one project or every project.
The full picture
Useful at every stage of the work.
Nine kinds of help, and in each one you set how much it does on its own.
Planning
Settle the work before coding starts.
- A plan you can comment on, section by section
- An overall design sketched first
- A project guide (wiki), kept current
Writing code
Write, improve and update your code.
- New features, fixes and tidy-ups
- Modernize old code, upgrade its libraries or translate it
- Comfortable with older code, COBOL included
Testing
Find problems before your customers do.
- Tests written for each change
- A self-check and test run at the end of each task
- A separate tester in every Orchestrator run
Reviewing and shipping
A second opinion, then a careful release.
- /review: as many as five reviewers together
- Each edit shown to you for sign-off, by default
- Asks first, then ships to its own branch
Knowing your code
Sees the whole project, not one file at a time.
- Cortex: a private map of how your code fits together
- Remembers past sessions and searches old chats
- Finds code by meaning, after a one-time setup
Working with your tools
Fits in with the systems your team uses today.
- Connections to your other tools and systems
- Tools and skills in one click from the marketplaces
- Optional: opens the app it built on your computer to test it
Keeping control
Fast, but inside the limits you set.
- Sign-off on by default, the Delegation Dial for everything else
- Allow, ask and deny rules, plus checks you write (hooks)
- An optional sandbox that fences commands in
- The Usage & Cost dashboard, with budget alerts
Bigger jobs
From a single chat to many helpers at once.
- Agent Manager: lots of chats on one screen
- Agent Teams of 2 to 5 that work together
- /batch: as many as 30 tasks, each in its own copy of the code
Everyday comfort
Small things that make the day easier.
- A save point at each step, restored in one click
- Optional typing suggestions and voice input
- Everyday English turned into a ready-to-run command
- Chat inside VS Code or in a browser tab
New lately
Fresh in Nx IDE.
- Teams
/batch, Agent Teams and Agent Manager
Run lots of AI helpers at once and let small groups work together. Break big changes into tasks that run side by side, each with its own draft pull request if you want.
- Plans
Reviewable plans, and /review
A plan you can comment on before anything gets built, and code reviews by as many as five reviewers together.
- Controls
Rules, hooks and a sandbox
Allow, ask and deny rules, checks you write for key moments (hooks), and an optional operating-system sandbox on macOS and Linux.
- Memory
Memory, research and cost tracking
Project memory across sessions, reading of web pages and documents, and the Usage & Cost dashboard with budget alerts.
- AI models
The latest AI models
New releases from the big AI providers, ready to choose in the model list.
Before you start
Default or optional?
Off until you switch it on
- The operating-system sandbox (macOS and Linux)
- Request and cost limits for each task
- Web searches and looking into other code projects
- Testing the app it built in a browser (needs Chrome or Edge)
- Automatic notes for the project memory
- Typing suggestions, read-aloud and voice input (voice input needs the FFmpeg tool and a speech-to-text key)
Set up once
- Finding code by meaning
- Tool connections: add your team's own, or install them from the marketplace
Can it act without asking us?
Only if you allow it.
Sign-off is on from the start, and the Delegation Dial plus your rules decide how much it may do alone. Turn the dial up and helpers started by another helper also run without asking, inside your deny rules.
Could costs get out of hand?
You see the estimated cost while it works.
Daily and monthly budget alerts warn you without stopping anything, and optional per-task limits pause the work until you decide.
Which AI can we use with it?
Any of 26 providers (21 in the TAA edition).
That includes models that run on your own computers, so your code never has to leave the building.
Does Nexgile ever see our code?
No.
Nx IDE sends us no usage reports (telemetry), and no Nexgile cloud sits in the middle. Your code goes only to the AI provider you picked.