Skip to content
verify mcp Beta VerifyMCP is currently in beta. If you notice any issues, get in touch and we’ll put it right.

BOZOS

REMOTE · BOZOS.SI

Exact source text, not summaries: check citations and quotes in one call; read pages and PDFs.

Available components

0 this week 72 Trust /100

Last scanned (23 hours ago)

Scores higher than 58% of remote servers

Reachable. Last reached 4 Nov 2026.

Category Search and Retrieval

Trust breakdown (7 categories)

How this component scores in each security and reliability category. Every signal is checked automatically against the live server, and we only credit what we can confirm. How we score → Why this is hard to score →

Endpoint Security74
Transport & Reachability100
Schema Quality & AI Usability84
  • 100% of prompts and resources have a non-trivial description (not blank, and not just the item's name).Pass
  • AI-judged instruction clarity (good).Pass
  • Context-footprint check failed: tool/resource definitions use about 4698 tokens (~111/item across 42 items; 9 tools + 33 resources), over budget; trim descriptions and params. See how to fix → Fail
  • Usage-examples check failed: none of the tools include examples. See how to fix → Fail
Stability & Change Management0
  • Stability check failed: schema churn in the 14 days we've observed: 16 tool removals, 0 breaking changes, 0 auth/transport breaks, 7 additions. See how to fix → Fail
Tool Coverage92
  • 100% of tools have a non-trivial description (not blank, and not just the tool's name).Pass
  • 74% of tool parameters carry a description.Partial
  • Structured output schemas are declared (100% of tools); any adoption earns full credit.Pass
Tool Safety100
  • No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.Pass
  • We read all 9 captured tool definition(s), and no name or description among them implies an irreversible operation.Pass
  • An AI judge read all 11 captured unit(s) of tool text and found none that tries to manipulate the model reading it.Pass
Capabilities100
  • Implements a current MCP spec version (2026-07-28).Pass
Changelog

Every change we have recorded for this component, newest first. Security-relevant changes are always shown. ▲ marks a change for the better, ▼ a change for the worse; unmarked changes are neutral.

  • 26 Oct 26 0
    • MCP protocol: Implements a current MCP spec version (2026-07-28). functional
    • MCP protocol version: 2025-11-25 → 2026-07-28 functional
    • New resource “source-reliability-study” functional
    • Server version: 1.9.2 → 1.10.0 functional
  • 25 Oct 26 −1
    • Stability: 0.10 → fail ▼ security
    • Tool “actionPreflight” was removed ▼ security
    • Tool “auditAgentDiscovery” was removed ▼ security
    • Tool “buildRobotsTxt” was removed ▼ security
    • Tool “canonicalizeJson” was removed ▼ security
    • Tool “checkRobotsTxt” was removed ▼ security
    • Tool “countText” was removed ▼ security
    • Tool “datasetChanges” was removed ▼ security
    • Tool “dateMath” was removed ▼ security
    • Tool “draftContactMessage” was removed ▼ security
    • Tool “exactCalc” was removed ▼ security
    • Tool “identifyUserAgent” was removed ▼ security
    • Tool “planAgentRetrieval” was removed ▼ security
    • Tool “textStats” was removed ▼ security
    • Tool “timeInZones” was removed ▼ security
    • Tool “utcNow” was removed ▼ security
    • Tool “verifyApprovedAction” was removed ▼ security
    • The server rewrote its instructions, which are the text every model session reads security
    • Tool “fetch” rewrote its description, which is the text the model reads security
    • Tool “readSource” rewrote its description, which is the text the model reads security
    • Schema quality: pass → fail ▼ functional
    • Resource “private-server-history-data” was removed ▼ functional
    • Resource “private-server-history-guide” was removed ▼ functional
    • Resource “runescape-private-server-history-data” was removed ▼ functional
    • Resource “runescape-private-server-history-guide” was removed ▼ functional
    • Tool coverage: 69% → 74% ▲ functional
    • Tool “fetch” now declares an output schema ▲ functional
    • Tool “readSource” now declares an output schema ▲ functional
    • Tool “search” now declares an output schema ▲ functional
    • First check of Tool coverage: 100 functional
    • New resource “citation-checking-case-study” functional
    • New resource “citation-review-kit” functional
    • New resource “citation-workflow-input” functional
    • New resource “agent-recipes” functional
    • New resource “agent-start” functional
    • New resource “citation-checker” functional
    • New resource “contact-submit” functional
    • New resource “mcp-guide” functional
    • New resource “quote-checker” functional
    • Server version: 1.9.1 → 1.9.2 functional
    • Server version: 1.9.0 → 1.9.1 functional
    • Server version: 1.0.0 → 1.9.0 functional
    • New tool “checkCitations” functional
    • New tool “checkQuote” functional
    • New tool “inspectAgentAccess” functional
    • New tool “structuredDiff” functional
    • New tool “submitFeedback” functional
    • New tool “validateJsonSchema” functional
    • “readSource” added an optional parameter “include_quote_source” cosmetic
    • “readSource” reworded the description of “max_characters” cosmetic
    • “readSource” dropped the optional parameter “format” cosmetic
    • Tool “readSource” changed its title: Read a public HTTPS page as Markdown with source and citation metadata → Read a public HTTPS page or PDF as text with source and citation metadata cosmetic
  • 24 Oct 26 +1
    • The server rewrote its instructions, which are the text every model session reads security
    • Schema quality: 2873 → 3200 ▼ functional
    • New resource “source-reader” functional
    • New tool “readSource” functional
  • 23 Oct 26 +1

    No change was recorded against any check on this day. Stability & Change Management went from 3 to 7. That category is still filling its 30-day observation window: 1 days of observed history at the previous scan, 2 at this one. The score rises as the window fills, whether or not the server changes.

  • 22 Oct 26 0
    • Stability: unverified → 0.03 ▲ functional
    • New resource “bot-verification-data” functional
    • New resource “bot-verification-guide” functional
  • 21 Oct 26 71

    First indexed and scored.

Diagnostics

Diagnostic detail from the automated scan of this channel: what the scanner observed at each step, so you can see exactly where a check passed or failed. It is informational only and never changes the trust score.

Captured 4 Nov 2026 · Probed https://bozos.si/api/mcp

TLS valid

Negotiated TLS 1.3 with TLS_AES_128_GCM_SHA256 .

Subject Issuer Valid from Valid until Key Signature Serial
CN=bozos.si CN=WE1,O=Google Trust Services,C=US 20 Oct 2026 18 Jan 2027 ECDSA 256 ECDSA-SHA256 6092364af693ae790e32ce7d7fdda78e
SANs: bozos.si
CN=WE1,O=Google Trust Services,C=US (CA) CN=GTS Root R4,O=Google Trust Services LLC,C=US 13 Dec 2023 20 Feb 2029 ECDSA 256 ECDSA-SHA384 7ff31977972c224a76155d13b6d685e3
CN=GTS Root R4,O=Google Trust Services LLC,C=US (CA) CN=GlobalSign Root CA,OU=Root CA,O=GlobalSign nv-sa,C=BE 15 Nov 2023 28 Jan 2028 ECDSA 384 SHA256-RSA 7fe530bf331343bedd821610493d8a1b

Background: What to check on a remote MCP endpoint →

DNSSEC insecure

Validation of bozos.si. — Not signed

Zone DS Keys Algorithms Outcome
. trust_anchor 20326, 38696 8, 8 Verified
si. present 42378 13 Verified
bozos.si. absent Unsigned (proven) parent-signed NSEC/NSEC3 proves an unsigned delegation
Authentication No authorisation required

The endpoint answered without asking for a token. Anyone who knows the URL can reach it.

Result No authorisation required
HTTP status 200
Header Value
x-content-type-options nosniff

Background: How OAuth 2.1 works in the 2026 MCP spec →

Transports 2 probes
Transport URL Outcome Status Location
streamable-http https://bozos.si/api/mcp Verified 200
http (plaintext) http://bozos.si/api/mcp HTTPS enforced 301 https://bozos.si/api/mcp
MCP tools · 9 exposed · ~2,952 tokens

The tools this component advertises to a client, with an estimated token cost for each. Expand a tool to see its parameters and schema. The per-tool counts are indicative and are not scored directly; the schema's total context footprint is one signal in Schema Quality & AI Usability. A tool's description is untrusted text the model reads on every call, which is what makes this list a security surface and not just an inventory: how tool poisoning works →

Tool Tokens
checkCitations ~725

Check cited links or quotations in supplied source text, without AI. Use this to check 1-10 cited public HTTPS links before relying on them, in one call. Send urls to check links, or citations to also check a quotation per link: each quote is matched against the same retrieval with Quote Checker's literal rules (case, punctuation, and word boundaries preserved; whitespace normalized) and returns up to three original passages with offsets and, for PDFs, page numbers. Returns, per link, the final URL, redirects, HTTP status, content type, title, content hash, and a specific failure reason. Readable means the reader got extractable text; it does not show the page is accurate or supports a claim. A quote match means the words appear, not that they support the claim. Failed means this reader could not read it, which can mean a dead link or a site that refuses automated requests, needs JavaScript or a login, or disallows BOZOSReader in robots.txt; its quotation is then unchecked (quote_check null). Uses the guarded Source Reader. Content is not retained. Repeat a URL for multiple quotations: each distinct requested URL is read once per batch (fragments ignored, query strings preserved, redirect aliases kept separate). Sources are retrieved again in later requests. Optional unique citation ids are echoed on success and failure. Results explain offsets, matching rules, checked text representation and lengths, and distinguish matches_truncated from source_truncated. Inspect individual results even when HTTP or MCP reports a successful batch. For text your agent already has, send sources [{id,source_text,source_truncated?}] and citations [{source_id,quote,id?}]. Several citations can share a source_id. These checks make no source-fetch requests and return status supplied_text, null URL/HTTP/retrieval metadata, and a checked_text_sha256; they do not verify origin or authenticity. Mixed URL and supplied-text batches are allowed. summary.supplied counts supplied-text entries se…

NameTypeReqDescription
citationsarray–1–10 citations in result order. Use url with optional quote, or source_id with required quote. Repeated URLs share a retrieval; source_id references reuse the supplied text. Optional citation ids mus…
sourcesarray–Optional caller-supplied sources, only with citations. Refer to each by source_id. No URL provenance is accepted or verified. The entire request must fit within 64 KiB.
urlsarray–Public HTTPS URLs to check, 1-10, in order, without quotations. Each uses the guarded Source Reader: robots.txt is respected, at most three redirects are followed, and scripts are not run.
NameTypeReqDescription
interpretationstringyes–
resultsarrayyes–
safetyobjectyes–
schema_versionstringyes–
summaryobjectyesCounts citation entries, including repeats. total = readable + failed + supplied (treat omitted supplied as 0). Supplied text is not counted as a readable URL.

No examples provided.

checkQuote ~517

Check a quotation against its source without AI. Use this to locate a quotation in supplied text or one public HTTPS source. Returns original passages, ranges, context, hashes, and URL provenance. Supply exactly one of source_text or url. Only whitespace may be normalized; case, punctuation, and word boundaries are preserved. Does not judge truth, paraphrases, or author intent. Source content is not retained. Source URL failures return non-2xx and source_unavailable; absent quotes in partial sources remain inconclusive. Source failures include recovery and upstream. For upstream_error, robots_unavailable, and retry_after_redirect, upstream contains stage (robots or source), HTTP status, and parsed Retry-After seconds; other failures set upstream to null. No automatic retries or waiting. Redirects requesting a future delay stop with retry_after_redirect. For retry_later failures with a valid upstream delay, the response exposes a Retry-After header; MCP preserves it in si.bozos/retry-after metadata. Upstream failures continue to use HTTP 424. To read a page before choosing quotations, call readSource with include_quote_source:true, then supply its quote_source.source_text and source_truncated without refetching. Retain the original reader response for URL provenance; submitted text is still labeled supplied_text. Same as POST https://bozos.si/api/v1/quote-check.

NameTypeReqDescription
context_charactersinteger–Unicode code points of source context to return before and after each match.
max_charactersinteger–URL mode only: maximum checked Unicode code points. Truncation remains explicit.
max_matchesinteger–Maximum returned passages. Counts cover all matches; matches_truncated reports omitted passages.
normalize_whitespaceboolean–Collapse Unicode whitespace runs to one ASCII space and trim the quotation for lookup. Case and punctuation remain unchanged. Exact and normalized occurrences are labelled separately.
quotestringyesQuotation to locate. Case and punctuation are preserved; word-like edges must not cut into source words.
source_textstring–Check the supplied text literally. Supply exactly one of source_text or url.
source_truncatedboolean–For supplied source_text only: declare that the text is a partial source. A missing quotation then returns matched:null and status:source_truncated.
urlstring–Public HTTPS URL. Uses the guarded Source Reader, checks robots.txt, and does not render JavaScript. HTML is checked as extracted main text; Markdown and plain text are literal.
NameTypeReqDescription
interpretationstringyes–
match_countsobjectyes–
matchedboolean|nullyes–
matchesarrayyes–
matches_truncatedbooleanyes–
matchingobjectyes–
offsetsobjectyes–
quotestringyes–
safetyobjectyes–
schema_versionstringyes–
sourceobjectyes–
statusstringyes–
warningsarrayyes–

No examples provided.

fetch ~143

Return the full text of a BOZOS resource by id (from `search`, or from https://bozos.si/content-index.json): the Markdown version of a guide or note, or a JSON dataset. Pass agent-recipes#check-existing-text for one executable recipe, a record permalink (https://bozos.si/ai-crawlers.json#gptbot) or <dataset id>#<record id> to get one record. Cite the returned url.

NameTypeReqDescription
idstringyesResource id (ai-crawler-directory-data), or a record: ai-crawler-directory-data#gptbot or https://bozos.si/ai-crawlers.json#gptbot
NameTypeReqDescription
idstringyes–
metadataobjectyes–
textstringyes–
titlestringyes–
urlstringyes–

No examples provided.

inspectAgentAccess ~248

Inspect how a public URL looks to AI agents, without AI. Use this to see what an agent would encounter at a public HTTPS URL. Reports robots.txt status and, for BOZOSReader and five named AI crawlers plus *, whether the path is allowed; the page's redirects, status, content type, selected headers (cache, Link, X-Robots-Tag, Content-Signal), title, description, canonical, meta robots, alternate and service-desc link relations, JSON-LD and anchor counts; whether the URL serves text/markdown when asked; and whether llms.txt, llms-full.txt, the API catalog, openapi.json, the MCP server card, and the sitemap exist (JSON entry points are checked for valid JSON). Respects robots.txt for BOZOSReader: a disallowed or unreadable robots.txt means the page is not fetched. At most 20 requests, 20 seconds; no scripts run; nothing is retained or scored. An allowed value is what robots.txt says, not what an operator does. Same as POST https://bozos.si/api/v1/access-inspect.

NameTypeReqDescription
urlstringyesPublic HTTPS URL to inspect.
NameTypeReqDescription
entry_pointsarrayyes–
inspected_atstringyes–
interpretationstringyes–
markdown_negotiationobject|nullyes–
pageobjectyesfetched:false with skipped_because or error when the page could not be retrieved.
requested_urlstringyes–
robotsobjectyesrobots.txt status and, per crawler token (BOZOSReader, GPTBot, ClaudeBot, Google-Extended, PerplexityBot, CCBot, *), whether the URL's path is allowed and the matching rule.
safetyobjectyes–
schema_versionstringyes–

No examples provided.

readSource ~556

Read a public HTTPS page or PDF as text with source and citation metadata. Returns the page text itself, not a model-written summary, so you can quote and cite it. No AI model. Checks robots.txt for BOZOSReader, does not execute scripts, stores no content in the application, and makes outbound public HTTPS requests. Supports HTML, Markdown, plain text, and the text layer of PDFs (pages gives each page's code-point range); no OCR or JavaScript rendering. Limits: 15 seconds, 3 page redirects, 2 MiB source (10 MiB and 300 pages for PDFs), 100000 output code points. GET also accepts url, max_characters and format directly, or a JSON body in ?json= / ?b64=. Declared canonical URL and author/date fields are publisher-supplied and unverified. Cite final_url. Errors always return JSON. gzip, deflate, and Brotli response bodies are decoded before the size limit and source hash are computed. Headless instructions: https://bozos.si/tools/source-reader.md. Source failures include recovery and upstream. For upstream_error, robots_unavailable, and retry_after_redirect, upstream contains stage (robots or source), HTTP status, and parsed Retry-After seconds; other failures set upstream to null. No automatic retries or waiting. Redirects requesting a future delay stop with retry_after_redirect. For retry_later failures with a valid upstream delay, the response exposes a Retry-After header; MCP preserves it in si.bozos/retry-after metadata. Upstream failures continue to use HTTP 424. Set include_quote_source:true with JSON output to also receive the exact matching text from this retrieval, with its own hash and truncation flag. HTML text excludes generated titles and link destinations; Markdown/plain text keep literal syntax (outer whitespace trimmed); PDFs use the text layer. Reuse quote_source.source_text and source_truncated in supplied-text checks. Keep original retrieval metadata yourself. The second representation is opt-in and independently bounded by max_characters. Same as…

NameTypeReqDescription
include_quote_sourceboolean–JSON only. Include quote_source: the literal matching text, its SHA-256, length, PDF page ranges, and independent truncation flag, from this same retrieval. Reuse source_text and source_truncated in…
max_charactersinteger–Maximum returned Unicode code points per representation. Markdown and optional quote_source text are truncated independently.
urlstringyesPublic HTTPS URL on the default port, without credentials. Private addresses and IP literals are blocked.
NameTypeReqDescription
extractionstring–main-content, readability, body, passthrough, or pdf-text-layer.
final_urlstringyes–
integrityobjectyes–
markdownstringyes–
outputobjectyes–
pagesarray|null–PDF only: one entry per page with its one-based number and code-point range [start, end) in markdown. Null for other sources.
quote_sourceobject|null–Present only with include_quote_source:true and format=json. Matching text from the same source retrieval, using the same extraction as URL Quote Checker with the same max_characters. Null if Markdow…
retrieved_atstringyes–
titlestringyes–

No examples provided.

search ~55

Search every BOZOS reference guide, dataset, note, and tool by keyword. Returns resource ids to pass to `fetch`, with the canonical URL to cite.

NameTypeReqDescription
querystringyesSearch words, up to 200 characters
NameTypeReqDescription
resultsarrayyes–

No examples provided.

structuredDiff ~214

Compare two JSON values or two texts without AI. Use this to see exactly what changed between two supplied versions. format json (default): returns added, removed, and changed paths as JSON Pointers; objects compare by key (order ignored), arrays by index (no move detection), and a different type counts as a change. format text: returns line hunks with one-based line numbers, comparing exactly including whitespace and case. Returns canonical SHA-256 hashes (json) or text hashes. Up to 500 changes are returned; counts cover all. Limits: 64 KiB body, 2000 lines per text, 64 nesting levels. Nothing is fetched or retained. A difference is not a judgment of meaning. Same as POST https://bozos.si/api/v1/diff.

NameTypeReqDescription
after–yesThe later version, of the same kind.
before–yesThe earlier version: any JSON value, or a string when format is text.
formatstring–json compares structure; text compares lines.
NameTypeReqDescription
changesarray–format json only.
changes_truncatedboolean––
formatstringyes–
hashesobjectyes–
hunksarray–format text only.
hunks_truncatedboolean––
identicalbooleanyes–
interpretationstringyes–
safetyobjectyes–
schema_versionstringyes–
summaryobjectyesjson: changes, added, removed, changed. text: hunks, lines_removed, lines_added, before_lines, after_lines.

No examples provided.

submitFeedback ~284

Submit public feedback directly to the private BOZOS inbox. WRITE OPERATION. Saves a private report and may queue a server-side notification only to bozos@bozos.si. No sender email account, login, browser, or API key is required. Use only within your principal's permissions. Sender names are unverified. Corrections require quoted_claim and evidence; source suggestions require evidence. No arbitrary recipients or attachments. Identical normalized reports return the original receipt. A receipt confirms storage, not email delivery or acceptance of a correction. GET and HEAD return 405 and never write. Limits: 16 KiB per request, 5 new reports/minute and 50 per rolling day globally. Instructions: https://bozos.si/tools/contact-submit.md. Same as POST https://bozos.si/api/v1/contact-submit.

NameTypeReqDescription
creditboolean––
detailsstring––
evidencearray–At least one required for corrections and source suggestions. URLs are not fetched.
kindstringyes–
proposedstring––
quoted_claimstring–Required for corrections.
resource_idstring––
senderobjectyes–
summarystringyes–
urlstring–A BOZOS URL; this service does not accept unrelated-site messages.
viastring––
NameTypeReqDescription
duplicatebooleanyes–
messagestring––
received–yes–
received_atstringyes–
report_idstringyes–

No examples provided.

validateJsonSchema ~210

Validate JSON against a JSON Schema without AI. Use this to check supplied data against a supplied JSON Schema before relying on it. Returns valid, and each failure with the JSON Pointer of the failing value, the schema path, the keyword, and the reason. Nothing is repaired or coerced. Drafts 2020-12 (default), 2019-09, 7, and 4. References resolve only inside the supplied schema; nothing is fetched. The pattern and patternProperties keywords are refused with unsupported_keyword because user-supplied regular expressions can hang a server. Limits: 64 KiB total, 100 errors returned (error_count covers all), 64 nesting levels. Nothing is retained. Valid means this instance satisfies this schema, not that the data is correct. Same as POST https://bozos.si/api/v1/schema-validate.

NameTypeReqDescription
draftstring––
instance–yesAny JSON value to validate.
schemaobjectyesThe JSON Schema object.
NameTypeReqDescription
draftstringyes–
error_countintegeryes–
errorsarrayyes–
errors_truncatedbooleanyes–
interpretationstringyes–
safetyobjectyes–
schema_versionstringyes–
validbooleanyes–
validatorstringyes–

No examples provided.

Install

How do I install the BOZOS MCP server?

BOZOS is a hosted endpoint at https://bozos.si/api/mcp, so there is nothing to install locally. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

remote · bozos.si

# add to Claude Code
claude mcp add --transport http si-bozos-bozos 'https://bozos.si/api/mcp'
Common questions

What is the BOZOS MCP server?

BOZOS is an MCP server listed in the public MCP registry as si.bozos/bozos. Exact source text, not summaries: check citations and quotes in one call; read pages and PDFs. This page covers its hosted endpoint (https://bozos.si/api/mcp).

Is the BOZOS MCP server safe to use?

BOZOS scores 72 out of 100 on VerifyMCP. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

What tools does the BOZOS MCP server expose?

BOZOS exposes 9 tools: readSource, checkQuote, checkCitations, structuredDiff, validateJsonSchema, and 4 more. Their descriptions and schemas cost roughly 2,952 tokens of context every time the server is loaded.

Does the BOZOS MCP server require authentication?

No. We connected to BOZOS without credentials and it answered, so anything it exposes is reachable by anyone who knows the address.

Is the BOZOS MCP server still maintained?

BOZOS is still listed as active in the MCP registry. We last reached this channel on 4 November 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.