What is Nx IDE?
The product, the Nexgile Code agent, and what it can do for you
Nx IDE is Nexgile’s AI-powered development environment. At its heart is Nexgile Code — the built-in AI coding agent (currently version 3.6.27) that lives in a panel beside your editor. You describe what you want in plain language; Nexgile Code reads your codebase, plans the work, edits files, runs commands, and reports back — while you review and approve every consequential step.
Nexgile Code is not an autocomplete widget. It is an agent: it carries out multi-step tasks end to end, shows its reasoning as it goes, and stops to ask you before it does anything that changes your project.
What Nexgile Code can do
Section titled “What Nexgile Code can do”| Capability | How it works |
|---|---|
| Read and understand your code | Reads whole files or line ranges, lists directories, searches by regex, and (optionally) searches by meaning with semantic codebase indexing. |
| Edit through reviewable diffs | Every proposed change appears as a diff you can inspect before it lands. Approved changes are also captured as checkpoints you can restore with one click. |
| Run commands | Executes terminal commands — builds, linters, scaffolding — with the output visible in chat, subject to command guardrails. |
| Drive your tests | Runs your existing test suites and reads the results; in Workflow Mode an independent Tester agent verifies work the implementer claims is done. |
| Connect to external systems | Speaks the Model Context Protocol (MCP) to reach issue trackers, browsers, databases, and internal services. See the MCP overview. |
Every one of these actions is approval-gated by default. Nothing touches your filesystem, terminal, or external services until you say yes — and you decide, category by category, what to auto-approve later.
Three ways to use it
Section titled “Three ways to use it”Most teams end up using all three patterns.
1. Single-mode work
Section titled “1. Single-mode work”Pick a persona — Code for implementation, Debug for diagnosis, Architect for planning, Ask for questions — and one agent handles the whole task. This is the everyday pattern: bug fixes, small features, “explain this module to me.” Learn how modes shape what the agent may do in Using Modes.
2. Workflow Mode
Section titled “2. Workflow Mode”An opt-in delivery chain for changes you would normally route through code review. An Orchestrator delegates to a fresh Code sub-agent that implements, an independent Tester that builds its own test plan and verifies, an independent Reviewer that reads the diff, and — only after you confirm — a Shipper that commits and pushes to a feature branch. Details in Workflow Mode.
3. Documentation engineering
Section titled “3. Documentation engineering”Two dedicated modes treat documentation as a product. Spec Bootstrap interviews you about a new project and generates a deeplink-strict wiki where every page is concrete, linkable, and verifiable. Spec Sync walks your git history after each sprint and surgically updates the wiki so it never drifts from the code.
The ten core modes at a glance
Section titled “The ten core modes at a glance”Modes are personas with defined permissions — the agent can only do what its current mode allows. The ten core modes:
| Mode | Best for | Can edit files? | Can run commands? |
|---|---|---|---|
| Architect | Planning, design, RFCs | Markdown only | No |
| Code | Implementation | Yes | Yes |
| Ask | Q&A, learning a codebase | No | No |
| Debug | Diagnosing & fixing bugs | Yes | Yes |
| Orchestrator | Coordinating multi-step work | No (delegates) | No (delegates) |
| Tester | Independent verification | No | Yes |
| Reviewer | Independent code review | No | Yes |
| Shipper | Stage, commit, push approved work | No | Yes |
| Spec Bootstrap | New-project documentation | Spec files only | Yes |
| Spec Sync | Keeping wikis current | Spec files only | Yes |
Beyond these, 21 specialist modes — Refactor, Test Gen, Doc Gen, Security, Translate, and more — cover focused jobs, and custom modes let you add your own.
What a task actually looks like
Section titled “What a task actually looks like”A concrete feel for the loop you’ll live in:
- You type a request — “Add input validation to the login form.”
- The agent asks to read the relevant files; you approve.
- It records a visible todo list of the concrete work items, then implements them in order.
- Each edit arrives as a diff; you read it and click Save (or Deny, with a note about what you’d prefer).
- After one verification pass, a Task Completed summary closes the loop — with the task’s token and dollar cost visible in the header the whole time.
The full walkthrough is Your First Task.
Why teams choose it
Section titled “Why teams choose it”- You stay in control. Every risky action goes through an approval prompt by default, and you choose exactly which categories to trust with auto-approval. Hard caps on per-task cost and request count act as circuit breakers.
- Free undo. Every agent-initiated change is checkpointed in a shadow git repository — separate from your project’s real git history — so any prior state is one click away.
- Bring your own AI provider. Connect Anthropic, OpenAI, Google Gemini, AWS Bedrock, GCP Vertex AI, OpenRouter, and many more — or run fully local with Ollama or LM Studio. No lock-in, no markup. See the providers overview.
- Project-aware. A
.nexgilerulesfile in your repo teaches Nexgile Code your conventions once, and every future task — including sub-agents — honors them. See Rules Files & AGENTS.md. - 31 modes out of the box. Ten core modes plus 21 specialist personas (Refactor, Test Gen, Security, Translate, DB Migrator, and more) are available from the mode dropdown with zero setup — and you can define your own.
Where to go next
Section titled “Where to go next”- Install and open it: Installing & Opening — Nexgile Code ships inside Nx IDE, so this takes minutes.
- Understand modes: Using Modes — the single biggest steering wheel you have.
- Learn by doing: Your First Task — the request → act → approve → complete loop, walked through end to end.