v2.0.0 — The Universal Context Compiler

Any source.
Any text box.

Channel the context. From messy local folders to web-based AI builders, dumpall is the fluid pipe that routes scattered code into one clean, pasteable stream.

> npx dumpall src/ -c
The flow state of text.

You have a folder. You need it in a text box. That's it.

dumpall isn't just an "AI prompting tool." It's a universal utility pipe. Whether you're feeding web-based builders like Lovable and Bolt, seeding a ClickUp doc for knowledge handoff, prepping context for a human code review, or writing a CI pipeline script—dumpall aggregates the noise into a single, predictable markdown file. No black magic.

01
The Arcs

Escaping the
walled gardens.

01
Vs. Bloated Packagers
(repomix, code2prompt)
  • Requires Python/Rust environments
  • Heavy config files
  • Assumes you're in a Git repo
The Dumpall Way Zero setup. npx ready. Works on raw URLs or local folders instantly. Pure CLI speed.
02
Vs. Walled IDEs
(Cursor, Windsurf)
  • Trapped in current workspace
  • Can't index remote GitHub repos
  • Hard to reference "that old project"
The Dumpall Way Cross-pollinate context. Run dumpall to grab files from an entirely different project without ruining your IDE's index.
03
Vs. Terminal Traps
(Claude Code, Copilot CLI)
  • Confined to terminal execution
  • Blind to external Web Docs
  • Useless for Web AI builders
The Dumpall Way Pipe it anywhere. Fetch a web sitemap, format to markdown, and pipe to clipboard to paste directly into Lovable, Bolt, or ChatGPT web.
02
Capabilities

Point at anything.
Exclude the noise.

@
Interactive Picker
Fuzzy-search and select files instantly. Works on local directories, URL sitemaps, and remote GitHub repos.
> dumpall @
URLs & Globs
Fetch any web page. Use globs to pull entire documentation sections via sitemap automatically.
> dumpall "https://docs.api.com/**"
</>
Remote Repos
Point at a GitHub/GitLab URL. Downloads, unpacks, and dumps. Repo zips are cached locally for speed.
> dumpall github.com/owner/repo
#
Token Budgets
Know your context size. Estimate tokens or aggressively truncate output to fit your exact LLM context window.
> dumpall . --max-tokens 50000
{ }
Format Agnostic
Markdown is default for chat, but structured XML and JSON outputs are built-in for strict API workflows.
> dumpall src/ --format xml
🔗
Share & Scan
Upload your dump instantly to paste.rs or catbox.moe. Generate a terminal QR code to pass context to mobile.
> dumpall src/ --share --qr
03
Reference

Zero ceremony.
Run it anywhere.

# Run without installing > npx dumpall src/ # Or install globally > npm install -g dumpall

One mental model.
Every flag.

# CORE I/O --clip, -c Copy output to clipboard --out, -o <file> Write output to file --note, -m [text] Prepend a prompt (omit value for interactive) --format <fmt> Output format: md (default), xml, json --stdin Read sources from stdin (one per line) --share Upload and print a shareable URL --qr Upload and display a scannable QR code # FILTERING --exclude, -e <pat> Exclude glob pattern --grep <pattern> Only include files containing pattern --max-file-size <size> Skip files larger than size (default: 1MB) # CONTEXT MANAGEMENT --tree-only Output only the directory tree --tokens Show estimated token count --max-tokens <n> Truncate to budget (drops largest files first) # SYSTEM & CACHE --follow-symlinks Follow symlinks during traversal --strict Abort on any error instead of skipping --no-cache Bypass cache for remote fetching --cache-ttl <dur> Cache TTL for branch refs (e.g. 1h, 30m, 0) --max-pages <n> Max pages to fetch for URL globs (default: 50) --install-completions Install shell completions (bash/zsh/fish)

Environment Variables.

DUMPALL_FORMAT Default output format (md, xml, json) DUMPALL_MAX_PAGES Default max pages for URL glob fetching DUMPALL_CACHE_DIR Override cache dir (default: ~/.cache/dumpall) # PRIVATE REPO AUTHENTICATION # Set token manually or let dumpall fallback to 'gh auth login' GITHUB_TOKEN Token for private GitHub repos GITLAB_TOKEN Token for private GitLab repos BITBUCKET_TOKEN Token for private Bitbucket repos
04
Workflows

In practice.

Feed AI your codebase
Grab source, exclude noise, clipboard ready.
# everything in src, skip tests, copy
dumpall src/ -e "*.test.ts" -c
Mix Everything
Local code, remote docs, and instructions—one paste.
# code + docs + interactive prompt
dumpall src/ https://docs.api.com/auth -m -c
Debug across files
Find every file touching an endpoint using inline grep.
# grep for the pattern, dump matches
dumpall . --grep "/api/users" -c
Pipe from Git
Only dump the files changed in your recent commits.
# last 3 commits → stdin → clipboard
git diff --name-only HEAD~3 | dumpall --stdin -c
Browse a private repo
Authenticate with token, open interactive picker.
# pass token inline, fuzzy search, copy
GITHUB_TOKEN=ghp_xxx dumpall github.com/user/priv @ -c
Grab docs via Glob
Use a wildcard to fetch a whole sitemap section.
# fetch all react hooks, format markdown
dumpall "https://docs.react.dev/reference/react/use*" -c