Documentation

Illumind docs

Based on aichat. All-in-one LLM CLI with chat, RAG, agents, function calling, and 20+ providers.

Getting Started

Request Access
First run & configuration
Environment variables

Usage

Command-line guide
Chat REPL
Shell assistant
Generate code
Use files & URLs
Run server

Features

Roles
Sessions
RAG
Agents
Macros
Function calling & MCP

Request Access

Submit your trial application at aichat.chat/trial. You'll receive an activation code within 24 hours. After activation:

$ aichat activate --code XXXX-XXXX

$ aichat --model claude-sonnet-4-7

$ aichat "review src/ for security issues"

First run & configuration

After receiving your activation code from the trial, Aichat creates a config file at:

Linux~/.config/illumind/config.yaml

macOS~/Library/Application Support/illumind/config.yaml

Windows%APPDATA%\illumind\config.yaml

Run aichat activate --code XXXX to activate with your trial code. After activation, you have full access to all models.

Command-line guide

$ illumind # Enter REPL

$ illumind Tell a joke # Generate response

$ illumind -e install nvim # Execute command

$ illumind -c fibonacci in js # Generate code

$ illumind -m openai:gpt-5 # Select model

$ illumind -r role1 # Use role

$ illumind -s session1 # Use session

$ illumind -a agent1 # Use agent

$ illumind --rag rag1 # Use RAG

$ illumind --serve # Run server

$ illumind -f image.png describe # Use files

Run illumind --help for full options list.

Chat REPL

Interactive chat with tab autocompletion, multi-line input, history search, and configurable keybindings.

Shell assistant

Describe what you want in plain English. Illumind generates and runs the right shell command for your OS.

$ illumind -e "find all .log files larger than 100MB"

Shell integration available for bash, zsh, PowerShell, fish, nushell. Press alt+e for inline completions.

Roles

Roles act as a prompt library. Create custom roles with system prompts for different tasks.

$ illumind --list-roles

$ illumind -r coder "fix the auth bug"

RAG

Feed documents, codebases, and URLs into conversations. Illumind chunks, embeds, and retrieves relevant context automatically.

$ illumind --rag myproject

$ illumind --rebuild-rag

Agents

Agents are like GPTs for your terminal. They combine instructions, tools, documents, and variables into reusable packages.

$ illumind -a security-audit

$ illumind --list-agents

Generate code

Use the -c flag to generate code without entering the REPL. Output is printed directly to stdout — pipe it into files or other tools.

$ illumind -c fibonacci in typescript

$ illumind -c 'express route for POST /users' > routes/users.ts

$ illumind -c 'dockerfile for node 20 app' | tee Dockerfile

Use files & URLs

Pass files, images, and URLs as context using the -f flag. Supports text, source code, PDFs, images, and remote URLs.

$ illumind -f src/auth.ts 'review this file'

$ illumind -f diagram.png 'describe this architecture'

$ illumind -f https://example.com/api/spec.yaml 'summarize the API'

$ illumind -f src/ 'find all functions without error handling'

Sessions

Sessions persist conversation history across invocations. Resume a session by name to continue a long debugging session or multi-step refactoring task.

$ illumind -s debug-auth # Start or resume session

$ illumind --list-sessions

$ illumind --delete-session debug-auth

Macros

Macros are reusable command sequences bound to a key or alias. Define them in your config file to automate multi-step workflows you run frequently.

# In ~/.config/illumind/config.yaml:

macros:

audit: '-a security-auditor -- review src/ for OWASP issues'

review: '-r senior-reviewer -f . summarize changes'

$ illumind :audit

$ illumind :review

Function calling & MCP

Connect LLMs to external tools, APIs, and data sources. Supports MCP (Model Context Protocol) for standardized tool integration.

Built-in server

Lightweight HTTP server with OpenAI-compatible API, LLM Playground, and Arena.

$ illumind --serve

LLM Playground

LLM Arena

Themes

Custom dark and light themes for syntax highlighting.

Configuration file

The config file is YAML. All settings are optional — Illumind works out of the box with defaults. Common settings:

modeldefault: claude-sonnet-4-7 (or any provider:model)

temperaturedefault: 0 — set higher for more creative output

save_sessiondefault: false — auto-save all sessions

highlightdefault: true — syntax highlight responses

light_themedefault: false — use light color theme

streamdefault: true — stream responses as they arrive

Full reference: run illumind --info to see all active configuration values and their sources.

Custom REPL prompt

Customize the REPL prompt string to show context like the active model, session name, or token count. Supports ANSI color codes and dynamic variables.

# In ~/.config/illumind/config.yaml:

repl_prelude: '({model}) '

# Result:

(claude-sonnet-4-7) ▶

Environment variables

API keys and provider credentials are set as environment variables. Illumind reads them automatically — no additional configuration needed.

OPENAI_API_KEYOpenAI / GPT-5

ANTHROPIC_API_KEYAnthropic / Claude

GEMINI_API_KEYGoogle / Gemini

GROQ_API_KEYGroq (fast inference)

MISTRAL_API_KEYMistral Large 3.1 AI

OPENROUTER_API_KEYOpenRouter (multi-provider)

LLM providers

Illumind supports 20+ providers out of the box:

OpenAIAnthropic (Claude)Google (Gemini)Mistral Large 3.1GroqOllamaAzure OpenAIAWS BedrockGoogle VertexCohereDeepSeekxAIOpenRouterTogetherPerplexityCloudflareReplicateErnieQianwenMoonshot