The CLI stack for building AI agents
A curated index of mature, scriptable command-line tools an agent can shell out to. Pick a category, or press ⌘K to search all 146.
Web & Research
HTTP clients, scraping, crawling, search, and DNS. How an agent reaches out to the internet and pulls structured data back.
Firecrawl · crawl4ai · monolith · pup
Media
Image, video, and audio: convert, transform, generate, transcribe, and download. The heavy binary work an agent should never attempt in-model.
FFmpeg · yt-dlp · gallery-dl · streamlink
Documents & OCR
Convert between formats, manipulate PDFs, pull text out of images and scans, and turn documents into clean markdown an LLM can read.
Pandoc · LibreOffice · Typst · mdcat
Data
Slice, filter, transform, and query structured data: JSON, YAML, XML, CSV, and SQL. The glue that turns raw output into something the agent can reason over.
jq · gojq · fx · jless
Development
Git, code search, static analysis, refactoring, testing, and benchmarking. The tools an agent uses to navigate and change a codebase.
git · GitHub CLI (gh) · git-cliff · gitleaks
Files & Cloud
Sync, transfer, back up, compress, and rename. Moving bytes between the agent's disk and cloud storage.
rclone · rsync · s5cmd · croc
Infrastructure
Containers, orchestration, infrastructure-as-code, secrets, and network diagnostics. How an agent builds, ships, and inspects running systems.
Docker · Docker Compose · Podman · kubectl
Modern CLI
Modern, fast, parseable replacements for classic unix tools. Better default output means less prompt space spent teaching the agent to read it, and cleaner data to pipe onward. (Search tools like ripgrep/fd live under [Development](development.md).)
bat · eza · hexyl · tldr
AI & Agent Utilities
The layer specific to running LLMs and agents: model CLIs, context builders, token accounting, sandboxes, and notifications. Tools an agent uses to call other models or prepare its own inputs.
llm · AIChat · Ollama · files-to-prompt