AI Coding Agent · v5.1.0 · Windows · macOS · Linux

Your terminal.
Your agent.
No limits.

FluxusCLI is a fully autonomous coding agent that runs in your terminal. It reads your codebase, writes production code, executes commands, diagnoses failures — and keeps going until the job is actually done.

Download Free Get started ↓
6
Providers
21
Tools
50
Auto loops
40+
Commands
Live preview
fluxuscli — ~/project
50
Autonomous loops
Runs self-directed until the task is complete — no babysitting required.
21
Built-in tools
Shell, files, search, diffs, HTTP requests, Python — everything an agent needs.
6
AI providers
OpenRouter, Anthropic, OpenAI, Groq, Cerebras, Mistral. Switch any time.
∅
Dependencies
Self-contained executable. No environment setup. No package manager.
01 · Features

Built for
real work.

Every capability was designed to solve an actual developer pain point — not to pad a feature list.

Scroll through the capabilities that make FluxusCLI the last tool you reach for.
01
Autonomous Agentic Loop
Runs up to 50 iterations without prompting. Reads code, writes code, runs commands, catches errors, and retries — independently until the job is done.
02
21 Integrated Tools
Shell execution, surgical file patching, regex search across the codebase, HTTP requests, Python evaluation, unified diffs, glob reads — all native, no configuration.
03
Six AI Providers
Switch between OpenRouter, Anthropic, OpenAI, Groq, Cerebras, and Mistral at any time — mid-session, no restart. Each provider's full model list is built in.
04
Persistent Project Memory
Use /remember to distil architecture decisions, bugs found, and codebase notes into memory that persists across every session automatically.
05
Deep Code Review
/review delivers a structured analysis: critical bugs, security flaws, design problems, and strengths — annotated with file names and line numbers.
06
Intelligent Git Commits
/commit reads the diff, generates a well-structured Conventional Commit message, stages all changes, and commits. No copy-pasting required.
07
Iterative Debugging
/debug <cmd> runs a failing command, reads the error, investigates the root cause across files, applies a fix, and re-runs to confirm. Repeats until resolved.
08
Named Sessions
Save full conversation state with /save and restore with /open. Your codebase understanding, decisions, and context never disappear between sessions.
09
Pipe & Script Mode
Pass a prompt via -p "task" or pipe from stdin for fully non-interactive use in CI/CD pipelines, scripts, and automation workflows.
10
Safe by Default
Destructive operations — delete, move — require explicit confirmation. API keys are written with 600 permissions. Toggle off with /noconfirm for scripts.
02 · Providers

Your key,
your model.

Connect any of six providers with a single command. Models are pre-loaded — no config files needed.

Use /key <provider> to connect. /models to browse. /models 1 to switch.
OpenRouter
Anthropic
OpenAI
Groq
Cerebras
Mistral
OpenRouter
Unified gateway · Access hundreds of models via one API key
Available models
stepfun/step-3.5-flashDefault · Fast
deepseek/deepseek-r1-0528Strong reasoning
deepseek/deepseek-chat-v3-0324Excellent coder
qwen/qwen3-235b-a22bTop-tier · Large
meta-llama/llama-4-maverickMeta flagship
mistralai/mistral-small-3.1-24b-instructCode · Fast
Anthropic
Claude model family · Best-in-class for agentic coding
Available models
claude-opus-4-6Most capable
claude-sonnet-4-6Speed + intelligence
claude-haiku-4-5Fastest
claude-sonnet-4-5Agentic
claude-opus-4-1Advanced coding
OpenAI
GPT and o-series · Broad capability across tasks
Available models
gpt-4oMultimodal flagship
gpt-4.1Smartest non-reasoning
gpt-4o-miniFast · Affordable
o3Deep reasoning
o4-miniFast reasoning
Groq
Ultra-fast inference · Ideal for rapid iteration
Available models
llama-3.3-70b-versatileAll-rounder
llama-3.1-8b-instant~560 tokens/s
deepseek-r1-distill-llama-70bFast reasoning
qwen/qwen3-32bStrong reasoning
Cerebras
High-throughput inference on custom silicon
Available models
llama-3.3-70bDefault
llama3.1-70bProduction
llama3.1-8bLightweight
Mistral
European AI · Excellent coding and reasoning models
Available models
mistral-large-latestTop-tier reasoning
codestral-latestCode specialist
magistral-medium-latestReasoning
mistral-small-latestEfficient
03 · Commands

Every command,
documented.

40+ commands, organized by purpose. Type /help inside FluxusCLI for the same reference in your terminal.

Select a category on the left to browse all available commands.
AI Workflows
Model
Files
Navigation
Shell
Session
Keys
Snippets
AI Workflows
Purpose-built for engineering tasks
/plan <task>Generate a numbered checklist before executing any work
/planstatusView progress of the active plan with completion tracking
/review [file]Deep code review: critical bugs, security issues, design flaws
/explain [file]Explain a file or codebase — architecture, flow, dependencies
/testAuto-detect the test suite, run it, and analyse results
/debug <cmd>Run failing command, diagnose root cause, fix it, verify
/commitAI-authored Conventional Commit message — stages and commits
/ask <question>Instant Q&A without tool calls — fast answers
/rememberDistil the session into persistent project memory
/memoryDisplay current project memory
/thinkToggle collapse/expand of AI reasoning sections
Model
Switch providers and models on the fly
/modelsList all models for the active provider
/models <provider>List models for a specific provider
/models allShow every model across all providers
/models <n>Switch to model number n from the active list
/modelShow current model ID and active provider
/model <id>Switch directly to any model by exact ID
/setmodelEnter and save a custom model ID interactively
/setmodel <provider>Save a custom model scoped to a provider
/list customList all saved custom models by provider
/delcustom <id>Delete a saved custom model
Files
Read, edit, copy, and manage files directly
/cat <file>View a file with syntax highlighting and line numbers
/add <file>Inject file contents into the conversation context
/grep <pattern>Search all files for a regex pattern with context lines
/diff <file>Colour-coded git diff for a specific file
/edit <file>Open the file in your $EDITOR
/touch <file>Create an empty file including parent dirs
/rm <path>Delete a file or directory with confirmation
/cp <src> <dst>Copy a file or directory
/mv <src> <dst>Move or rename a file or directory
Navigation
Move around your project structure
/cd <path>Change the agent's working directory
/pwdPrint the current working directory
/ls [path]List directory contents with sizes and timestamps
/tree [path]Render a recursive directory tree
/find <pattern>Fuzzy filename search across the working directory
Shell
Execute commands and inspect the environment
!<cmd>Execute a shell command directly
/run <cmd>Run a shell command (same as ! prefix)
/watch [N] <cmd>Re-run a command every N seconds, print changes
/py <code>Execute a Python snippet inline
/profile <file>Profile a script and show performance hotspots
/depsScan imports and verify installed vs missing packages
/gitGit status with branch and modified files
/envPrint all environment variables
/which <cmd>Find a binary's path in PATH
Session
Manage history, state, and configuration
/clearClear the entire conversation history
/historyPrint all messages in the current session
/undoRemove the last exchange from history
/redoResend the last message
/retryRetry the last message fresh
/compactSummarise history while preserving key facts
/save [name]Save the current session under a given name
/open [name]Load a saved session or list all sessions
/rename <old> <new>Rename a saved session
/statsSession duration, turns, tool calls, token usage
/costToken usage breakdown for this session
/config [k=v]Show config or set a persistent key=value
/noconfirmToggle confirmations for destructive operations
Keys & Providers
Manage API credentials securely
/keyShow active provider and all saved key statuses
/key <provider>Securely enter and save an API key for a provider
/key providersList all providers with key and active status
/key removeDelete all saved API keys
Snippets
Save and reuse AI-generated content
/snippetList all saved snippets with metadata
/snippet save <name>Save the last AI response as a named snippet
/snippet <name>Insert a saved snippet into the conversation
/snippet del <name>Delete a named snippet
04 · Install

Running in
two minutes.

Download the executable, run it, connect a provider, and start building. No environment setup, no package manager.

Works on Windows, macOS, and Linux out of the box.
01
Download
Click Download Free above. Extract the zip anywhere on your machine. No installer needed.
02
Run FluxusCLI
Launch FluxusCLI.exe (Windows) or ./FluxusCLI (macOS/Linux) and point it at your project folder.
03
Connect a provider
Type /key openrouter and paste your API key. OpenRouter gives access to dozens of models with a single key.
04
Pick a model and go
Type /models to browse, /models 1 to select. Then type any task in plain English — FluxusCLI handles the rest.
Windows
# run from the extracted folder .\FluxusCLI.exe C:\projects\myapp
macOS / Linux
chmod +x FluxusCLI ./FluxusCLI ~/projects/myapp
Connect and select a model
❯ /key openrouter ❯ /models ❯ /models 1
Script / pipe mode
./FluxusCLI -p "refactor auth.py to use JWT"   echo "add tests for api.py" | ./FluxusCLI
Stop waiting.
Start building.
Version5.1.0
PlatformsWindows · macOS · Linux
Providers6 supported
Tools21 built-in
DependenciesNone
LicenseOpen Source
AuthorOrangepeels24
Download FluxusCLI View install guide ↑