BOOT · STARTING
A free, local-first terminal for building precise AI prompts with the
Action → Role → Context → Response protocol. The third tool in the OWL family,
after OWL (agent) and knOOgle (search). Every step supports
auto-suggest — type auto or pick ★ auto and owl: arcr
uses the previous step plus a live research fan-out to fill in a sensible value.
Final prompts are emitted as JGF (JSON Graph Format) with sources, keywords,
definitions, and the composed natural-language prompt embedded. Everything runs in
your browser: no account, no tracking, no server, and nothing leaves the tab except
the queries you send to the public APIs you enable.
ACTION — the verb that tells the AI what to do (compose, analyze, build…).
ROLE — the perspective or expertise the AI should adopt (engineer, ethicist…).
CONTEXT — the scenario, audience, constraints, and details. Must be specific.
RESPONSE — the deliverable format (report, code, image, diagram…).
Every step supports auto or ★ auto. Auto-suggest uses the
previous step and the live research engine to fill a sensible value. At the CONTEXT
step, auto fans out across the research sources and synthesises a
context from the top result. The final build always attaches a sources list, even
if you skipped research — owl: arcr fans out on your context keywords before
emitting the JGF.
Every ARCR prompt is emitted as JGF (JSON Graph Format) — a structured graph
with nodes for action, role, context,
response, keywords, sources, definitions, visual/agentic/contract
metadata, plus edges describing their relationships
(adopts, applies_to, delivers,
supported_by, cites, governed_by, …).
JGF is both machine-parseable and human-readable. The composed natural-language
prompt lives in metadata.prompt so agents that prefer prose can still
use it. Every final output has COPY and EXPORT .JSON buttons.
owl: arcr fans your context out across keyless public APIs — Wikipedia,
Wikidata, OpenAlex, arXiv, Europe PMC, Crossref, GitHub, Hacker News, Stack Overflow,
npm, Lobsters, Dev.to, Open Library, Datamuse, DuckDuckGo (opt-in) — scores each
result by keyword overlap, freshness, depth and source weight, and attaches the top
references as source nodes in the JGF. Toggle each source from the terminal with
/sources.
/help full command list /arcr [seed] start an ARCR session (optional topic seed) /skill build a reusable SKILL.md /research <q> fan out across the research engine /dict [term] self-dictionary lookup /dict add t = d add a dictionary term /export [json|skill] export last prompt (JGF) or skill /history saved prompts /agentic toggle agentic execution loop (default: on) /sources list research sources /stats local stats /about version and credits /reset reset session (keeps history + dictionary) /clear /cancel clear screen / cancel flow
owl: arcr is agent-friendly. Build a prompt entirely from URL parameters:
?action=build&role=engineer&context=design+a+renewable+energy+system&response=code ?action=...&role=...&context=...&response=...&format=json ?research=quantum+computing ?research=quantum+computing&format=json ?file=llms.txt ?file=agent-search.json
Serve the discovery files at the standard paths — /llms.txt and
/.well-known/agent-search.json — to satisfy bot discovery. The app serves
them at ?file=llms.txt and ?file=agent-search.json as a
fallback.
Everything stays in your browser. There is no account, no cookie, no analytics, no
telemetry. The only outbound traffic is the query text sent to the public APIs you
have enabled, plus any webhook you explicitly configure. No OpenAI, Anthropic, or
Google endpoints are ever contacted. Your prompts, dictionary, and saved history
live in localStorage on this device and are never uploaded.
owl: arcr is a free tool from webXOS. If it saves you time, you can buy the developer a coffee: buymeacoffee.com/webxos ☕. Donations are optional and never unlock anything extra — the tool is and stays free.