Skip to content

Features

Everything you need to orchestrate AI development.

A full VS Code-compatible editor with an agent, a planner, connectors and memory built in. Five pillars, one window.

01 · Agent

An agent that does the work, where you can see it.

Agent mode works across the whole folder: it reads, edits, runs commands and checks its own results — and you review every change before it lands.

  • Up to 25 turns per task, with a live turn counter and cost ticker
  • Commands run in a visible Nexora Agent terminal, behind a security denylist
  • Four run modes, from ask-every-time to run-everything
  • Every edit arrives as an Accept/Reject diff — or auto-apply when you trust the run
  • Mention files and folders with @, stop any turn instantly
  • Inline code suggestions at the cursor (early)
Nexora Agent · turn 12 / 25 · $0.018
✓ read src/app/checkout/page.tsx
✓ edited src/lib/stripe.ts +31 −2
✓ ran npm run test — 58 passed
› 3 files waiting for your review
▍

02 · Orchestration

One request becomes a plan you can read, price and approve.

Nexora classifies what you asked for, breaks it into a dependency graph and assigns the right platform to each step. Nothing live happens until you approve.

  • Dependency-graph execution with parallel steps, retries and timeouts
  • Estimated cost per step before running, actual cost after
  • Approvals survive a restart; edit, re-route or cancel any step
  • Built-in templates, save any plan as a template, share as .nexflow
  • Workflow graph view with live progress over WebSocket
  • Execute mode for trusted, pre-approved flows
plan p_7f3 · 5 steps · est. $0.04
① scaffold landing page Claude $0.03
② create products table Supabase free
③ add checkout Stripe free
④ open pull request GitHub free
⑤ deploy preview Vercel free
▍

03 · Memory

Most tools remember your code. Nexora also remembers your decisions.

Context is engineered in layers, so Nexora stops forgetting what you said yesterday — and all of it stays on your machine.

  • Recent turns of every session replay into the next prompt
  • A searchable archive of past chats the agent can query
  • Standing project facts that every chat and agent turn can recall
  • Workspace memory in a single portable local file
  • Run history with suggestions drawn from what worked before
  • Stored under your user profile — never on our servers
memory · acme-store
fact stack: Next.js 14 + Supabase + Stripe
fact deploys go to preview before production
chat 2 Oct · chose Stripe Checkout over Elements
run 3 Oct · landing-page deploy · $0.03 · ok
▍

04 · Connectors

Your platforms, connected once and used as steps.

Sign in once and Nexora can use a platform inside any plan — or hand a task to another coding agent and bring its work back for review.

  • 11 live connectors: OpenAI, Claude, GitHub, Vercel, Supabase, Stripe, Tavily, ElevenLabs and more
  • 45 platforms catalogued in the registry
  • Connect vendor-hosted MCP servers by signing in; 8 registered out of the box
  • Run Claude Code, Cursor or Cline in an isolated git worktree, accept file by file
  • Nexora Browser: the agent's page streamed live inside the IDE
  • Agent-to-agent card for delegating work between systems
connections
● GitHub signed in · OAuth
● Vercel signed in · OAuth
● Stripe test key · SecretStorage
● Claude Code worktree agent · ready
○ Supabase connect →
▍

05 · Control

Autonomy you can audit, and undo.

Nexora is built on the idea that agents should earn trust step by step. You decide what runs, see what it cost, and can reverse what it did.

  • Human approval before deploys, pull requests and payments
  • Rollback of recorded actions on external platforms
  • Time-travel timeline of every run and decision
  • Analytics for spend, runs and outcomes
  • Bring your own keys — no markup on model usage
  • Local engine on loopback with a per-install token
timeline · today
14:02 plan approved by you
14:03 PR #42 opened
14:04 preview deployed
14:09 rollback requested → preview removed
▍