Nexora Code

Professional multi-model AI coding agent & terminal.
Built for engineers who don't want to leave the terminal.

📖 Documentation · 🐛 Report a Bug · 💡 Request a Feature

What is Nexora Code?

Nexora Code is a fully open-source, terminal-native AI coding agent that brings Gemini-style conversational intelligence directly into your workflow. It is built on a BYOK (Bring Your Own Key) model — you connect your own API keys from providers like Gemini, Anthropic, OpenAI, OpenRouter, and Groq, and Nexora Code handles everything else.

Unlike browser-based AI tools, Nexora Code lives entirely in your terminal:

• Multi-action agent — reads, writes, refactors and fixes files autonomously
• Conversation memory — sessions persist across restarts
• Smart auto-routing — picks the right model for code vs. general tasks
• 50+ built-in tools — file ops, Git integration, AST mapping, semantic search, web fetch, test runner, and more
• Multi-provider BYOK — swap models on the fly without reconfiguring
• Session history — browse and resume past conversations with /chats
• Background tasks — run long agent jobs non-interactively via --headless
• OS Keychain integration — API keys stored securely in your native keychain, never in plaintext

Install via npm

The fastest way to get started. Requires Node.js v20+:

npm install -g nexora-code

Or run directly without installing:

npx nexora-code

Nexora Code will guide you through connecting your first API key on first launch.

Quick Start

# Start interactive chat
nexora

# Ask the agent to write or modify code
nexora write "build a REST API in Express with user auth"

# Fix a file
nexora fix src/components/Button.tsx

# Index your project into memory
nexora scan

# Run a one-shot prompt (CI/scripting)
echo "Summarise this file" | nexora --headless

Supported Providers

Nexora Code is model-agnostic. Connect any combination of these providers:

| Provider | Models | |----------------|---------------------------------------------| | Gemini | Gemini 2.0 Flash, Gemini 1.5 Pro, and more | | Anthropic | Claude 3.5 Sonnet, Claude 3 Haiku, and more| | OpenAI | GPT-4o, GPT-4o Mini, o1, and more | | OpenRouter | 200+ models via a single key | | Groq | Llama 3.1, Mixtral, Gemma (ultra-fast) |

Manage keys at any time:

nexora auth login gemini
nexora auth list
nexora auth remove openai

Commands

CLI (one-off)

| Command | Description | |----------------------------|------------------------------------------------------| | nexora | Start interactive Chat mode | | nexora scan | Index current project folder into semantic memory | | nexora write <desc> | Generate files from a description (Agent mode) | | nexora fix <file> | Analyse and fix a file autonomously (Agent mode) | | nexora auth login <p> | Add or rotate an API key for a provider | | nexora auth list | List all configured providers | | nexora auth remove <p> | Remove a provider's API key | | nexora doctor | System health: keys, DB, cache, bootstrap state | | nexora stats | Token usage, cache hit rate, cost estimates | | nexora debug | Internal paths, config state, runtime diagnostics | | nexora task create <job> | Spawn a background agent task | | nexora task list | Show running background tasks | | nexora reset | Factory reset: wipe config, memory, and API keys | | nexora uninstall | Clean local data and guide through full removal | | nexora --headless | Pipe a prompt via stdin for scripting / CI use | | nexora --version | Show version | | nexora --help | Full command reference |

Inside the REPL

| Command | Description | |----------------------|------------------------------------------------| | /new | Start a blank new conversation | | /chats | Browse and resume past conversations | | /chat models | List & select a model for Chat mode | | /agent models | View & change Agent models | | /auto | Toggle smart auto-model routing | | /mode agent | Switch to Agent mode (file read/write) | | /mode chat | Return to Chat mode | | /mode auto | Agent auto-approves all trusted writes | | /mode plan | Agent previews full plan before executing | | /scan | Index current project into semantic memory | | /write <desc> | Generate or create files from a description | | /fix <file> | Analyse a file for bugs and patch it | | /resume | Resume the last interrupted agent session | | /history | Recent agent sessions and their status | | /setup | Re-run the API key setup wizard | | /auth | Manage API keys (login, list, remove) | | /doctor | Show system health report | | /stats | Token usage and cost stats | | /debug | Internal paths and config state | | /task create <job> | Start a background task | | /task list | Show all running background tasks | | /task watch | Stream output from a running task | | /task result | View the output of a completed task | | /task cancel | Cancel a running background task | | /clear | Redraw the full UI | | /help | Show command reference | | exit | Quit Nexora Code |

Uninstall

Using the CLI (Recommended)

nexora uninstall

This removes local configuration, sessions, and API keys, then prints the final npm command to complete removal.

npm uninstall -g nexora-code

Modes:

| Mode | What it removes | |-----------------------------|----------------------------------------------------| | nexora uninstall | ~/.nexora, keychain tokens | | nexora uninstall --soft | Config & keys only, preserves logs & cache | | nexora uninstall --deep | Complete wipe of all local Nexora data | | nexora reset | Wipes config & memory, keeps the binary installed |

Project Structure

nexora-code/
├── src/
│   ├── index.ts                ← Entry point & CLI router
│   ├── types.ts                ← Global type definitions
│   ├── agent/                  ← Agent engine, tools, session, safety
│   │   ├── engine.ts           ← Multi-action agent loop
│   │   ├── tools.ts            ← 50+ built-in agent tools
│   │   ├── autonomy-engine.ts  ← Fully autonomous task execution
│   │   ├── safety-engine.ts    ← Rollback, sandboxing, approval gates
│   │   ├── semantic-rag.ts     ← Retrieval-augmented generation
│   │   ├── vector-search.ts    ← Semantic project indexing
│   │   ├── procedural-memory.ts← Memory for long-term task contexts
│   │   ├── git-tools.ts        ← Native Git integration tools
│   │   ├── scaffold-tools.ts   ← Boilerplate and template generation
│   │   ├── deploy-tools.ts     ← Deployment automation tools
│   │   ├── test-tools.ts       ← Automated testing and validation
│   │   └── subagents.ts        ← Swarm subagent orchestration
│   ├── commands/               ← CLI subcommands
│   │   ├── chat.ts             ← REPL loop, memory, streaming
│   │   ├── auth.ts             ← API key management
│   │   ├── init.ts             ← Project initialization wizard
│   │   ├── tasks.ts            ← Background task manager
│   │   ├── doctor.ts           ← Diagnostics & health checks
│   │   └── uninstall.ts        ← Uninstallation wizard
│   ├── config/                 ← Configuration files
│   │   ├── agent-config.ts     ← Agent behavior settings
│   │   └── chat-config.ts      ← Chat UI and model configurations
│   ├── core/                   ← Infrastructure layer
│   │   ├── provider-registry.ts← Multi-provider BYOK routing
│   │   ├── context-manager.ts  ← Token budget & context windowing
│   │   ├── circuit-breaker.ts  ← Provider failover & retries
│   │   ├── ast-mapper.ts       ← Abstract Syntax Tree analysis
│   │   ├── telemetry.ts        ← Performance and metric tracking
│   │   └── sandbox.ts          ← Isolated execution environment
│   └── utils/                  ← Shared utilities
│       ├── bootstrap.ts        ← Runtime environment validation
│       ├── keychain.ts         ← OS keychain integration
│       ├── mcp-client.ts       ← Model Context Protocol client
│       ├── multi-project.ts    ← Multi-workspace management
│       ├── learning-system.ts  ← Adaptive learning from feedback
│       ├── router.ts           ← Smart auto-model routing logic
│       ├── db.ts               ← SQLite database connection
│       └── ui.ts               ← Terminal UI chrome & animations
├── packaging/                  ← Platform install/uninstall scripts
├── .github/workflows/          ← CI/CD pipelines
└── package.json

Architecture Highlights

• ESM-native TypeScript — fully typed, no legacy JavaScript baggage
• Single bundled binary — esbuild bundles everything into dist/bundle.cjs so installation is a single npm command
• OS Keychain security — API keys are stored in macOS Keychain, Windows Credential Manager, or Linux Secret Service via keytar. Never written to disk in plaintext.
• SQLite + Vector Search — better-sqlite3 + sqlite-vss powers semantic code search, session history, and caching
• Provider Circuit Breaker — automatic failover to backup providers if a primary provider times out or errors
• Safety Gates — all destructive agent actions (file writes, deletes, shell commands) require explicit confirmation unless running in --auto mode

Contributing

Contributions are very welcome! Nexora Code is MIT-licensed and community-first.

Good first issues include:

• Adding new agent tools
• Improving provider support (new models, APIs)
• Enhancing the terminal UI
• Writing tests and docs

To get started:

git clone https://github.com/enochboadu/nexora-code
cd nexora-code
npm install
npm start          # Run in development mode

For larger changes, please open an issue first to discuss your idea. We review PRs actively.

Roadmap

• [ ] Cloud session sync across machines
• [ ] MCP (Model Context Protocol) full integration
• [ ] Multi-agent orchestration (spawn sub-agents for parallel tasks)
• [ ] Plugin marketplace for community tools
• [ ] Nexora Cloud — hosted model gateway with no BYOK required
• [ ] VS Code extension with shared context

Follow development at lab.getnexora.cloud and star the repo to stay updated.

Data Storage

All Nexora Code data is stored under ~/.nexora/ on all platforms:

| File / Directory | Contents | |------------------------|---------------------------------------| | config.json | Preferences, model selection | | state.json | Bootstrap and setup state | | nexora.db | Session history, semantic index | | sessions/ | Conversation session archives | | logs/ | Debug logs | | cache/ | Response cache |

API keys are not stored in ~/.nexora. They live exclusively in your OS keychain.

License

MIT © Nexora Inc. — Founded 2026 by Enoch Boadu

Nexora Code is the open-source core of the Nexora developer platform. The goal is to make professional AI-assisted development accessible to every engineer on every team — regardless of which AI provider they prefer or how much they want to spend.