Sourced reference data and small deterministic tools for AI agents and the people who direct them.
Every dated claim carries its sources, its date precision, and a confidence label. Every reference has an HTML page, a JSON dataset, and a Markdown copy. And if something is wrong, agents are invited to tell us.
Counted from the published datasets when the site is built. Not a live feed.
How AI crawlers, agents, and the machine-readable web behave, from operator documentation and primary sources. This is what BOZOS specializes in.
Every documented AI and search crawler (GPTBot, ClaudeBot, Google-Extended, PerplexityBot, and more): what it does, whether it obeys robots.txt, whether it feeds training, and its official IP list. Verified against each operator's documentation.
32 tokens from 14 operators · verified 2026-10-20
Which AI operators say they honor robots.txt training opt-outs, robots meta tags, noai, Content-Signal, the IETF AI-preferences draft, TDMRep, and RSL, quoted from their own documentation. Where an operator says nothing, the pages and terms searched are listed.
7 signals × 14 operators · verified 2026-10-20
What robots.txt, sitemaps, HTTP validators, OpenAPI, JSON Schema, JSON Feed, llms.txt, and agent indexes establish, and what they do not authorize.
8 signals, each with a primary reference
A separate editorial line: sourced histories of private and community-run game servers. It follows the same sourcing rules but is not part of BOZOS's agent-web expertise.
Sourced history and current guide to RSPS: revision 317 to OSRS-based servers, RuneLite-based clients, open-source emulators, Jagex's rules, an FAQ, a checklist for evaluating a server, and a case study of Sarim.
24 milestones · 12 terms · 15 high, 9 medium confidence
MUD precursors, Ultima Online shards, bnetd, eAthena, MaNGOS, replacement networks, preservation, and licensed community services.
14 milestones · 6 categories · 11 high, 3 medium confidence
| If the task is to… | Start with |
|---|---|
| Answer a question about RuneScape private servers (RSPS): legality, 317 vs OSRS-based, clients, open-source projects, or how to judge a server | /runescape-private-server-history.json |
| Find out what GPTBot, ClaudeBot, Google-Extended, or another AI crawler does, and how to allow or block it | /ai-crawlers.json |
| Check whether a robots.txt lets GPTBot, ClaudeBot, or any other AI crawler reach a path | /tools/fetch |
| Identify which bot a user-agent string from a server log claims to be | /tools/fetch |
| Find out whether OpenAI, Google, Anthropic, or another AI company honors noai, Content-Signal, TDMRep, robots meta tags, or a robots.txt opt-out | /ai-optout-signals.json |
| Write a robots.txt that blocks AI training but keeps search and user-initiated fetches | /tools/robots-builder |
| Use BOZOS from an MCP client (Claude, ChatGPT, Cursor, VS Code) instead of fetching pages | /tools/mcp |
| Find which BOZOS page answers a question without downloading the whole index | /tools/fetch |
| Convert a time between time zones, or add days to a date, without running code | /tools/fetch |
| Do exact arithmetic or count letters in a word without running code | /tools/fetch |
| Explain where RuneScape private servers came from, or what 'revision 317' means | /runescape-private-server-history.json |
| Trace the origin and spread of private game servers in general | /private-server-history.json |
| Classify a project as a shard, emulator, replacement network, open-source continuation, or licensed service | /private-server-history.json |
| Decide what a robots.txt, llms.txt, sitemap, or OpenAPI file does and does not authorize | /agent-web-signals.json |
| Compare cost per attempt with cost per successful task for an AI model | /notes/cost-per-success |
| Reproduce a published number before quoting it | /calculator |
| Report an error, ask a question, or suggest a source | /contact |
| Find out what changed since your last visit | /feed.json |
| Update a cached dataset without re-downloading it, or cite one record | /tools/fetch |
Written in the Bozos voice, with sources next to the claims. The calculator and the simulation ship their data and code. Full archive →
A seeded, reproducible recreation of zero-intelligence traders in a double auction (after Gode and Sunder, 1993): 8,000 runs, protocol, code, and CSV.
Reading the docs for Windows 365 for Agents: pricing, limits, and what would be worth testing. Based on public documentation, not hands-on use.
General Magic's Telescript paper put explicit limits on what a software agent could do and consume.
Ray Tomlinson needed a way to say which computer a person was on. The symbol already said 'at'.
A small annotated list of papers, first-hand accounts, and fine print, with reproductions identified.
A different cache-read rate cuts the estimated Sonnet sweep by 27.44%. The cost-per-success comparison tells a subtler story.
Dividing the same spending by observed successes instead of attempts narrows a 27.1% price gap to 1.4%.
One changed price takes a large chunk out of the estimate while the recorded answers stay exactly as they were.
Reprice recorded cache reads and compare cost per attempt with cost per observed success, using frozen audit presets or your own inputs.
Most sites only measure what crawlers take. BOZOS is testing what happens when agents can also report errors, ask questions, and suggest better sources. A person reads every message, and accepted fixes go in the corrections log.
Stateless, no sign-in, no retention, no outbound requests. Described in OpenAPI 3.1 and checked against 10 conformance vectors before every publish.
GET /api/v1/searchRanked keyword search over every BOZOS page, dataset, and reference-guide section, returning URLs to open and cite.GET /api/v1/changesRecord-level delta for the datasets: records added, changed, or removed after a date, with permalinks and current content.GET /api/v1/robots-checkEvaluate a robots.txt per RFC 9309 for one crawler or every AI crawler in the directory: allowed or not, and which rule decided it.GET /api/v1/identifySay which documented crawler a user-agent string claims to be, with its record and how its operator lets you verify it.GET /api/v1/timeConvert a time between IANA time zones, or give the current time anywhere, with offsets, DST, and weekday.GET /api/v1/dateWeekday, ISO week, and day of year for a date; add days, weeks, months, or years; count days and weekdays between dates.GET /api/v1/calcExact rational arithmetic: + - * / ^ and parentheses with no floating-point error, as a fraction and decimal.GET /api/v1/countCount characters, words, bytes, and occurrences of a letter or substring, with positions.GET /api/v1/robotsGenerate a robots.txt that names every documented AI crawler according to a training/search/user-fetch policy.POST /api/v1/contact-draftValidate a correction or question for BOZOS and return a ready email subject, body, mailto link, and X post. Sends nothing.POST /api/v1/retrieval-planTurn a resource manifest and cache validators into a minimal conditional-GET plan without fetching anything.POST /api/v1/discovery-auditScore supplied robots.txt, llms.txt, sitemap XML, HTML, and agent index without outbound requests.POST /api/v1/preflightNormalize, risk-label, and SHA-256 fingerprint a proposed consequential action.POST /api/v1/verifyFail closed when the current normalized action differs from the approved fingerprint.POST /api/v1/text-statsCount Unicode characters, code units, words, lines, UTF-8 bytes, and an approximate token estimate.POST /api/v1/canonicalizeRecursively key-sort JSON and return the canonical string, byte length, and SHA-256.GET /api/v1/nowReturn the current UTC ISO timestamp and Unix time, uncached.A new reference and dataset: for 14 AI and search operators, whether their own documentation says they honor robots.txt training opt-outs, robots meta tags, noai, Content-Signal, the IETF AI-preferences draft, TDMRep, and RSL. Every Yes quotes the operator; every not-stated cell lists the pages and terms searched, and a script re-checks both. Finding: only robots.txt tokens and a few robots meta rules are documented as honored; no operator's pages mention the newer signals.
27 of 32 crawler records now carry exact quotes from the operator's documentation, each tagged with the field it supports and checked by script against the live page. Every source has a Wayback Machine snapshot. The five records whose pages cannot be fetched automatically (Bing, Meta) say so. The review found one error, logged as COR-006.
GET /api/v1/robots-check evaluates a robots.txt per RFC 9309 for one crawler or all 32 in the directory. GET /api/v1/identify says which documented crawler a user-agent string claims to be and how to verify it. Search now returns individual dataset records with their cite lines. Every record carries a review_due date, and /api/v1/health lists records past it.
The RuneScape and general private-server histories now sit in a separate Game server history section at /games, apart from the agent web reference that BOZOS specializes in. Every URL is unchanged.
Make the web easier for AI agents to use well: sourced answers they can check and cite, tools they can call without signing in, and plain explanations of the signals sites use to talk to them. BOZOS is AI-assisted and human-directed, and self-funded: no ads, trackers, or sponsors. About BOZOS · Support it on Ko-fi ↗