only-cli

Turn most website into a command line interface, so AI agents like Claude Code, Codex, and Antigravity can browse without burning tokens on raw HTML or screenshots.

A typical page is tens of thousands of tokens of markup. The signal on it fits in a few hundred. only-cli fetches the page, distills it into a compact text view with numbered actions, and lets an agent drive the site by number:

$ oc open news.ycombinator.com
# Hacker News
[1] Show HN: I built a tiny CSV toolkit
[2] 312 comments
...
actions: do <n> | read <n> | next | raw

$ oc do 1

No per-site adapters required, no browser extension, no daemon. One generic distillation engine, three runtime dependencies, and a hard token budget on everything it prints.

Install

npm install -g @only-cli/oc

Requires Node 20+. Requests go through impers impersonating Chrome; if impers is unavailable the tool falls back to native fetch.

For AI agents

The fastest setup is one line in your agent's instructions file (CLAUDE.md, AGENTS.md, or equivalent):

When you need content from a web page, run npx @only-cli/oc open <url> instead of fetching raw HTML. Run npx @only-cli/oc --help once to learn the commands.

Claude Code users can install the skill instead: copy skills/only-cli/ into your project's .claude/skills/ directory (or ~/.claude/skills/ to enable it everywhere). A skill costs almost no tokens until the agent actually invokes it, which fits how this whole project thinks. The same skill also installs through the skills.sh directory into Claude Code, Cursor, Codex, Copilot, and others:

npx skills add only-cli/oc

No setup at all also works: npx @only-cli/oc runs without a global install, and the tool teaches its own command surface through --help, the actions: line at the bottom of every render, and error messages that name the next command to run.

Commands

oc open <url>          fetch and render a page with numbered actions
oc do <n>              follow the numbered link [n] from the last page
oc next                the next budget worth of the page already open
oc read <n>            full text of the region at [n], up to 2000 tokens
oc raw [url]           distilled markdown of the whole page
oc fill <n> <text>     type into a numbered input               (v0.2)
oc submit [n]          submit a form                            (v0.2)

Flags: --budget <tokens> (default 500, 2000 for read), --json, --html (raw as cleaned HTML instead of markdown), --session <name> (separate page state per name), --verbose/-v (metrics on stderr: tokens saved vs the page HTML, HTTP status and client identity, timing, transfer size, memory; alias --stats, or export OC_VERBOSE=1). Metrics are off by default because they cost tokens too; agents should pass --verbose only when running verbosely.

oc open remembers the page it rendered, so oc do 3 follows [3] without the agent ever handling a URL. That state is a JSON file per session under ~/.only-cli (override the directory with OC_HOME); there is no daemon and no background browser. Links that search engines wrap in a tracking redirect resolve to the real destination, so do on a result works like a click.

Reading past the budget

A 500 token view of a long page is cheap on the first read and expensive on the second, if the only way to see more is the whole page again. So the view says what it left behind, and there are two ways to collect it that are not the whole page:

$ oc open https://old.reddit.com/r/linuxquestions/comments/xpznb1/best_terminal_web_browser/
...
[80] Kind of a weird question, but does anyone here use a terminal browser to ... +312 chars
... 570 more blocks (~4,362 tokens): 'oc next' for the next ~500, 'oc raw' for all
actions: do <n> | read <n> | next | raw

$ oc next     # the next ~455 tokens, continuing exactly where the view stopped
$ oc read 80  # that one comment in full, 88 tokens
$ oc raw      # the whole thread, 9,670 tokens

Neither next nor read fetches anything: open saves the distilled page, so continuing to read it costs one file read. read <n> takes one region, which is the block at [n] with the couple of blocks that lead into it, or the whole section when [n] is a heading. Headings and text blocks long enough to be cut get numbers for exactly this reason, and ... +312 chars on a line is how much of that block the view did not print.

Supported websites

only-cli works on any mostly-static website with no per-site setup: news sites, blogs, documentation, forums, search engines. One generic engine distills whatever HTML comes back. On top of that, clis/ ships tuned command shortcuts for:

website	domain	shortcuts
Hacker News	news.ycombinator.com	top, new, item <id>, user <name>
Reddit	reddit.com (via old.reddit.com)	sub <name>, post <id>, user <name>, search <query>
GitHub	github.com	repo <owner> <name>, user <name>, search <query>, trending, issues <owner> <name>
LinkedIn	linkedin.com	profile <name>, company <name>, jobs <query> (public guest views)
DuckDuckGo	duckduckgo.com	search <query>, lite <query>
Bing	bing.com	search <query>, news <query>
Stack Overflow	stackoverflow.com (via Atom feeds)	question <id>, tag <name>, user <id>, recent

The engine also renders Atom and RSS feeds as regular pages. That is how Stack Overflow works: the site serves every HTML page a Cloudflare challenge, but publishes full question and answer bodies under /feeds, so oc open stackoverflow.com/feeds/question/11227809 returns the question and its top answers in about 500 tokens. The same trick applies to any site that gates its pages but leaves its feeds open.

Not supported yet: pages that only render with JavaScript (a headless fallback is planned for v0.3), sites behind logins (page state is saved, cookies are not, so logins land in v0.2), and sites with hard bot challenges that do not expose feeds. Adding a site shortcut is a small JSON file; see CONTRIBUTING.md.

Benchmarks

Measured against live sites in only-cli/benchmarks (only-cli 0.2.0-beta.1, 2026-08-18). What each method hands the agent per page view, across six real tasks (an article, a news front page, a Reddit discussion, a web search, a GitHub repository search, a LinkedIn company page):

method	success	total tokens	avg ms
oc open	6/6	1,936	540
oc raw	6/6	21,334	541
OpenAI computer use (screenshot floor)	6/6	4,590
Claude computer use (screenshot floor)	6/6	6,294	840
Browser Use (state message)	6/6	6,470	2,543
Jina Reader	6/6	16,402	636
lynx -dump	5/6	24,657	457
Playwright MCP (snapshot)	6/6	25,832	365
Playwright rendered HTML	6/6	101,361	730
Selenium rendered HTML	6/6	166,557	1,189
raw HTML fetch	6/6	177,685	406

The compact view reads all six pages for less than half the tokens of its cheapest rival, a single-screenshot floor, and 92x fewer than raw HTML. The nearest rivals are floors, not full reads: the computer-use rows price a single 1024x768 screenshot, one look at the top of the page, and Browser Use's state message drops most page text. Among methods that deliver the page content, the gap is 8x to Jina Reader and 13x to Playwright MCP's accessibility snapshot. oc was also the only cleaner to return real content on all six tasks: lynx and the naive fetcher hit a DuckDuckGo challenge, and Reddit served Jina its block page. Jina Reader is also the only method in the table that routes browsing through a third party: every URL the agent reads is sent to Jina's servers, while oc talks only to the site itself.

The end-to-end agent benchmark runs Claude Code headless (claude -p on claude-sonnet-5) on six live tasks, one web tool per session, and reads success, turns, tokens, and cost from its JSON output. Three tasks read a single page; three start on one page and must follow a link to a second:

tool	success	turns	total tokens	total cost USD	avg s
oc	6/6 ✅	31	871,909	0.74	13 ✅
lynx -dump	6/6 ✅	29 ✅	772,831 ✅	0.55 ✅	14
raw curl	4/6	61	1,031,894	0.54	39
Jina Reader	6/6 ✅	30	855,243	0.72	19
Playwright MCP	6/6 ✅	48	1,575,695	1.22	29

The ✅ marks the best value per column among tools that finished every task.

Every token claude billed per tool across the six tasks, failed runs included:

oc             ###################                        871,909 tokens  31 turns
raw-curl       ########################################  1,855,550 tokens  61 turns  2 failed
lynx           #################                          772,831 tokens  29 turns
jina-reader    ##################                         855,243 tokens  30 turns
playwright-mcp ##################################        1,575,695 tokens  48 turns

Each session gets a skill documenting its tool, so every condition runs at its best. Two results are worth stating plainly. oc and lynx were the only tools that returned real content on every task: Reddit served Jina Reader and Playwright MCP a 403, so both "answered" that task by reporting the block, and raw curl burned its full turn budget there and on the GitHub search, roughly 400k tokens each, and returned nothing. But lynx, not oc, took the token and cost columns this round, and the reason was a missing feature. The compact view leaves link URLs out to save tokens, and the version under test had no way to follow one, so an agent had to re-fetch the page as oc open --json or oc raw just to learn where [15] points, while lynx -dump prints a references list for free. That tax is most of the gap on the multi-step tasks. oc do <n> has since landed and closes it: following a Hacker News story into its comments now costs about 3k characters instead of the roughly 23k the re-fetch route spent. The next benchmark run will say whether that is enough to take the column.

The same six tasks run through OpenAI's Codex CLI (codex exec) as well, where oc comes out ahead: of the two tools that returned real content on all six tasks, oc spent 287,862 tokens and lynx 408,548, and Playwright MCP spent 699,810 on the Reddit thread alone. Codex's per-session overhead is much smaller than Claude Code's, so the two agents are compared within their own tables rather than against each other. The benchmark repo has per-task numbers, per-tier breakdowns, methodology, and instructions for adding other tools and models.

Status

Early. v0.1 covers static pages, budget-aware rendering, and offline tests. Sessions, oc do <n>, oc next, and oc read <n> are in, the rest of the actions (fill, submit, find, back) land in v0.2, and a lazy headless fallback for script-heavy pages in v0.3. The design principles and how to contribute are in CONTRIBUTING.md.

Known limits, honestly: no JavaScript rendering yet, no sites behind logins yet, and pages behind hard bot challenges may still refuse the tool.

Contributors

• only-cli, creator and maintainer

License

MIT