cliara 0.5.0

AI shell assistant for real terminal work.
Natural language, macros, smart git/deploy helpers, and safer execution flow.

Why • Before/After • Install • Quick Start • Workflows • MCP • Safety • Docs

Cliara wraps your existing shell (bash, zsh, PowerShell, cmd).

• Normal commands still run as-is.
• ? turns intent into shell commands.
• Macros turn repeated command chains into reusable commands.
• push, deploy, and ? fix speed up common dev loops.

No new terminal to learn. You keep your shell habits and get an AI layer on top.

Why Cliara

Most AI terminal tools are great in demos and weak in daily workflows. Cliara is built for repetitive, error-prone, real development loops.

Before / After

The point is not replacing shell skills. The point is reducing repetitive glue work.

Install

pip install cliara

Run

cliara

First launch opens GitHub sign-in once for cloud features. After that, run cliara normally.

Quick Start (60 seconds)

cliara ~/projects/myapp ❯ ? kill whatever is using port 3000
✓ cliara ~/projects/myapp ❯ ma deploy-prod
✓ cliara [deploy-prod] ~/projects/myapp ❯ push
✓ cliara [deploy-prod] ~/projects/myapp ❯ deploy
X 1 cliara [deploy-prod] ~/projects/myapp ❯ ? fix
cliara [deploy-prod] ~/projects/myapp ❯ help

Core Workflows

1) Natural language to commands

cliara ~/myapp ❯ ? show largest files in this folder
cliara ~/myapp ❯ ? find when I last changed docker config
cliara ~/myapp ❯ ? explain this command output

2) Macros for repeated routines

cliara ~/myapp ❯ ma release-check
cliara ~/myapp ❯ mc
cliara ~/myapp ❯ release-check
cliara ~/myapp ❯ ml

3) Built-in helpers for shipping

cliara ~/myapp ❯ push
cliara ~/myapp ❯ deploy
cliara ~/myapp ❯ ? fix

4) Ask your codebase (RAG)

Index every git-tracked file into a local vector store, then ask questions in plain English and get answers with file:line citations.

cliara ~/myapp ❯ index
✓ Indexed 126 file(s), 1045 chunk(s).
cliara ~/myapp ❯ ? how does auth work
cliara ~/myapp ❯ ask where is the config loaded

• index builds (and incrementally refreshes) the index — only changed files are re-embedded; index rebuild, index status, and index clear are also available.
• ask <question> always searches the code; ? <question> does too once an index exists and the question looks code-related.
• The store is a local SQLite database under ~/.cliara/codebase_index/; nothing leaves your machine except the embedding/answer calls to your configured LLM.

5) Give your AI tools Cliara's context (MCP)

Cliara runs as an MCP server, so Claude Code, Claude Desktop, and Cursor can use what Cliara knows about your machine: the codebase index, the commands you've actually run, your saved macros, and the safety gate.

pip install 'cliara[mcp]'
claude mcp add cliara -- cliara mcp

Or one entry in any MCP client's config:

{
  "mcpServers": {
    "cliara": { "command": "cliara", "args": ["mcp"] }
  }
}

No API keys in the config — the server reuses the credentials you already set up.

Your assistant can then search your codebase semantically, look up how you did something last time, check a command's risk before running it, review a diff, and reuse your macros instead of reinventing them. It is read-only by default; add --allow-exec to let it run commands through Cliara's gate.

See docs/MCP.md for the full tool list and safety model.

Safety By Default

Cliara is designed to help you move fast without blindly executing dangerous commands.

• Risky commands trigger stronger confirmation flow.
• Potentially destructive operations can show a diff preview before execution.
• You can still inspect and control what gets run.

Who It Fits

• Developers who live in terminal all day
• Teams with repeated setup/release/debug routines
• People who want shell speed without command memorization overhead

Docs

• Quick Start
• MCP Server
• Cliara Cloud Deployment
• Postgres Setup

Troubleshooting

If cliara is not recognized:

python -m cliara.main

License

MIT