btlabs Core
Server Details
btlabs Core MCP endpoint: site content and brand data for AI agents. OAuth or API key.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 11 tools
Each tool maps to a clearly separate content type (pages, posts, faqs, glossary, media, tags, team, testimonials, branding), so an agent can pick by resource. The main risk is the near-identical shape of the editorial collections (findPages vs findPosts vs findTestimonials all read as 'find + richText content'), which only the long descriptions disambiguate. Also, the 'find' prefix obscures that these tools also create, update and delete.
All names are camelCase with a verb+object shape, which is readable and mostly uniform (findX, getWritingGuidelines, reportContentIssue). However, the dominant 'find' verb actively misdescribes the tools, which are full CRUD endpoints per their own descriptions — a naming pattern that predicts the wrong capability. 'brand' is also referenced as a separate tool that does not exist, muddying the pattern.
11 tools is a well-scoped size for a headless-CMS surface with roughly nine collections plus two cross-cutting helpers. Each listed tool corresponds to a real content domain; nothing obvious is redundant. The count only looks low because several tools the descriptions depend on are absent rather than because the present ones are excessive.
The descriptions repeatedly instruct the agent to call tools that are not in the set: schedulePost (scheduling publishing), listAvailableBlocks (required before writing any page or post layout), getPersona (the only way to read a team member's writing voice), addMediaFromUrl (the only way to add media binaries), and getResourceSchema. The pages tool also says to link new pages in the 'menus' collection, for which no tool exists, leaving navigation a dead end. Category/taxonomy collections (faq-categories, categories) are likewise referenced as relations but cannot be managed.
Publisher details
- Operator
- Berger+Team · Publisher source
- Operator website
- https://www.berger.team · Publisher source
- Vendor relationship
- First-party · Publisher source
- Documentation
- https://btlabs.dev/.well-known/mcp.json · Publisher source
- Privacy policy
- Not provided
- Trust center
- Not available
- Restrictions
- API key required. Keys are issued by the operator on request and are scoped per tool; there is no public self-service signup. A read-only key is available for directory health checks. · Publisher source
Available Tools
11 toolsfindBrandingAInspect
Read this site-wide settings record (one document, no id).
Single source of truth for brand identity, voice, business facts and site identity — READ FIRST when drafting content or answering about the business. Groups: voice (voiceTone, forbiddenTerms, allowedVariants, terminology) · business identity (businessType, served areas, availabilityStatus, pricing — when priceItems exist, quote them and never answer "price on request") · contact + emergencyContact · visual assets (logoLight/logoDark/favicon/defaultOgImage = upload IDs; ready-made URLs in /brand.json as logo_url/og_image_url/…) · imageStyleProfiles = the image-style canon, read it BEFORE generating images (scoping/fallback and prompt guidance are described on the fields; resolve reference-image binaries via brand find/get) · specDocs = brand/styleguide DOCUMENTS as bare upload IDs (logo files, a UI-spec HTML, web-font files). The IDs alone say nothing — resolve them via brand find/get to read filename, description and URL, and consult them when you need the visual canon (spacing, type scale, component look) that the token list does not cover. Editor usage rules often sit in the asset description ("NICHT auf weissem Hintergrund"). · htmlElements snippet templates (ready-made, brand-styled HTML/Markdown to copy and adapt — check here before hand-rolling markup) · aiAgentSystemPrompt (public bot persona, served at /brand.txt + /brand.json) · newsletter signup form copy. Field-level meaning: call getResourceSchema "branding" — fields carry editor descriptions. UPDATE POLICY (human-curated SSoT): change ONLY the specific field(s) you were explicitly asked to change — never bulk-rewrite, never "improve" neighboring fields. Many fields are localized: write per target locale, sequentially (one locale per call). branding global (singleton). Public read: yes via /api/agents/global. MCP capabilities: find, update. Fields: logoLight (upload: brand|media), logoDark (upload: brand|media), logoSquare (upload: brand|media), favicon (upload: brand|media), defaultOgImage (upload: media|brand), separator, specDocs (upload: brand), imageStyleProfiles (array), htmlElements (array), addressing (enum: du (You (informal))|sie (You (formal))), companyName, legalName, ownDomains, siteName (localized, required), tagline (localized), allowedVariants (textarea), responsiblePerson, founders (rel-array: team), shortDescription (localized), longDescription (richText, localized), vision (textarea, localized), values (textarea, localized), targetAudience (textarea, localized), usps (textarea, localized), notOffered (textarea, localized), voiceTone (enum-array: professional|casual|matter-of-fact|heartfelt|technical|clear|+21), voiceDescription (textarea, localized), forbiddenTerms (textarea, localized), exampleQuotes (textarea, localized), terminology (textarea, localized), fixedTerms (textarea), writingRules (textarea), industry (localized), businessType (enum-array: Electrician|Plumber|HVACBusiness|HousePainter|RoofingContractor (Roofing Contractor (roofer, roof repair))|GeneralContractor|+63), foundingYear (number), teamSize (localized), location (localized), softwareApplication (group), serviceRadiusKm (number, ≥0), servedAreas (textarea, localized), servedPostalCodes (textarea), availabilityStatus (enum: accepting (Accepting new clients)|limited (Limited capacity)|waitlist|notAccepting), availabilityNote (textarea, localized), priceItems (array), priceRange, priceNote (textarea, localized), brands (textarea), partners (textarea), certificates (textarea, localized), spokenLanguages (textarea), brandOs (richText, localized), vatId, taxId, chamberOfCommerce, reaNumber, shareCapital (localized), sdiCode, pecEmail, hostingProvider (textarea, localized), mailProvider (textarea, localized), storageProvider (textarea, localized), paymentMethods (localized), paymentTerms (number), jurisdiction (localized), websiteCredit (group), email, phone, address (textarea, localized), emergencyContact (group), geoLatitude (number, -90–90), geoLongitude (number, -180–180), bookingUrl, bookingCtaText (localized), bookingNote (localized), whatsappUrl, contactMessage (textarea, localized), socialLinks (textarea), contactPage (rel: pages), aboutPage (rel: pages), businessHoursKind (enum: onsite (Open on site)|phone (Reachable by phone)), weekly (array), businessHoursNote (textarea, localized), formsSubmitDefault (localized), formsSubmitting (localized), formsSelectPlaceholder (localized), formsThankYouHeadline (localized), formsThankYouBody (textarea, localized), formsGenericError (localized), enabled (boolean), checkboxLabel (localized), pendingNote (localized), alreadyActiveNote (localized), aiAgentSystemPrompt (textarea), aiCatalogQueries (textarea, localized), aiSeoRules (textarea), internalLinkRules (array), internalLinkExclusions (rel-array: pages|glossary|jobs|posts|team), internalLinkMaxSuggestions (number, 1–20).
Inputs: depth, locale, select, populate, +1 more.
findFaqsCInspect
FAQ entries. Both question (text) and answer (richText) required. Link to faq-categories via categories: [id]. Used by /faq system page + faqBlock. Has drafts: use _status:"published" to go live (draft stays hidden); schedule a future date via the schedulePost tool (collection "faqs"). faqs collection. Drafts: yes (publish workflow; a find WITHOUT draft:true returns the published main-table state only). Public read: yes via /api/agents. MCP capabilities: find, create, update, delete. ⚠ delete is PERMANENT (hard-delete; the admin Trash does NOT catch MCP deletes). Write ONE document at a time: wait for the response of each create/update/delete call before sending the next — parallel writes can fail with a database deadlock (HTTP 500, nothing saved; retry that call alone). Before writing public text here, call getWritingGuidelines with this collection and the locale (brand voice, writing rules, fixed terms, SEO rules, taken keywords, field rules). Fields: question (localized, required), answer (richText, localized, required), categories (rel-array: faq-categories), publishedAt (date), searchText (textarea, localized, ≤60000 chars, redacted).
Inputs: id, page, sort, depth, +10 more.
findGlossaryBInspect
Public glossary of terms. title required in default-locale; synonyms (comma-separated) and title are both used for content auto-linking. shortDefinition (required, max 300 chars) is the citable, self-contained blurb shown in tooltips and picked up by AI answers — keep it independently understandable. definition is optional Lexical richText for a longer explanation. related links to other glossary entries; noAutolink: true opts a term out of automatic in-content linking. glossary collection. Drafts: yes (publish workflow; a find WITHOUT draft:true returns the published main-table state only). Public read: yes via /api/agents. MCP capabilities: find, create, update, delete. ⚠ delete is PERMANENT (hard-delete; the admin Trash does NOT catch MCP deletes). Write ONE document at a time: wait for the response of each create/update/delete call before sending the next — parallel writes can fail with a database deadlock (HTTP 500, nothing saved; retry that call alone). Before writing public text here, call getWritingGuidelines with this collection and the locale (brand voice, writing rules, fixed terms, SEO rules, taken keywords, field rules). Fields: title (localized), slug (localized), synonyms (localized), shortDefinition (textarea, localized, required, ≤300 chars), definition (richText, localized), publishedAt (date), tags (rel-array: tags), related (rel-array: glossary), relatedFaqs (rel-array: faqs), noAutolink (boolean), robots (enum: index (Index (default))|noindex (noindex (hide))), searchText (textarea, localized, ≤60000 chars, redacted), meta (group: title, description, focusKeyword, secondaryKeyword1-4, searchIntent, eeatInsight, image — localized).
Inputs: id, page, sort, depth, +10 more.
findMediaAInspect
Content images, videos and documents (stock + uploads). MCP can READ + UPDATE metadata (alt-text, copyright, aiOrigin = "generated"/"edited" for AI images — set it for every image from an AI generator; it is written into all image sizes and shown in the credits; aiNotice = true additionally shows the EU "AI" icon on the image or video wherever it appears: use it for AI images/videos that look like real recordings of something real or plausible, not for drawings or stylised illustrations; never set by the system — when unsure, leave it out and tell the user) + DELETE (⚠ PERMANENT — the admin Trash does NOT catch MCP deletes); binaries come via the addMediaFromUrl tool, not raw upload (admin-only). Use id when referencing in upload-fields. VIDEOS: subtitles (per locale) = the id of a WebVTT (.vtt, text/vtt) media item — import the .vtt via addMediaFromUrl (url or upload: true), then updateMedia the video with "locale"; every block showing the video renders the captions in the page language. media collection. Drafts: no. Public read: yes via /api/agents. MCP capabilities: find, update, delete. ⚠ delete is PERMANENT (hard-delete; the admin Trash does NOT catch MCP deletes). Write ONE document at a time: wait for the response of each create/update/delete call before sending the next — parallel writes can fail with a database deadlock (HTTP 500, nothing saved; retry that call alone). Before writing public text here, call getWritingGuidelines with this collection and the locale (brand voice, writing rules, fixed terms, SEO rules, taken keywords, field rules). Fields: alt (localized, required), title (localized), caption (localized), description (textarea, localized), aiAutofillStatus (enum: pending|done|error (AI generation error)|skipped, redacted), poster (rel: media), subtitles (rel: media, localized), copyrightHolder, copyrightLicense (enum: all-rights-reserved|own-work|stock-licensed (Licensed (stock))|cc0 (CC0 (Public Domain))|cc-by|cc-by-sa|+2), copyrightUrl, aiOrigin (enum: generated (Created with AI)|edited (Edited with AI)), aiNotice (boolean).
Inputs: id, page, sort, depth, +10 more.
findPagesCInspect
Editor-curated marketing pages. Call listAvailableBlocks first to learn the available layout-block types. Title is required in default-locale. Use _status="draft" until reviewed, then _status="published" to go live NOW. To publish (or unpublish) at a FUTURE date, use the schedulePost tool with publishAt/unpublishAt — do NOT set a future publishedAt (the page stays a draft until the scheduled time, then auto-publishes). HOMEPAGE convention: set slug="home" (hardcoded in frontend route, EVERY locale needs slug="home"). System-pages use systemPageKey field — read-only in MCP. After page creation, link them in menus collection (slug="header-main") so they appear in navigation. INCOMING LINKS while unpublished: only link an anchor phrase that already exists in a published text (no new sentence, no rewording; new referring sentences only on publication day). pages collection. Drafts: yes (publish workflow; a find WITHOUT draft:true returns the published main-table state only). Public read: yes via /api/agents. MCP capabilities: find, create, update, delete. ⚠ delete is PERMANENT (hard-delete; the admin Trash does NOT catch MCP deletes). Write ONE document at a time: wait for the response of each create/update/delete call before sending the next — parallel writes can fail with a database deadlock (HTTP 500, nothing saved; retry that call alone). Before writing public text here, call getWritingGuidelines with this collection and the locale (brand voice, writing rules, fixed terms, SEO rules, taken keywords, field rules). Fields: title (localized), slug (localized), layout (blocks, localized), systemPageKey (enum: faq (FAQ index)|search (Search results)|posts-archive|glossary-archive|not-found (404 — page not found)|server-error (500 — server error)|+7), publishedAt (date), robots (enum: index (Index (default))|noindex (noindex (hide))), searchText (textarea, localized, ≤60000 chars, redacted), meta (group: title, description, focusKeyword, secondaryKeyword1-4, searchIntent, eeatInsight, image — localized). Layout blocks (39): hero, trustMarquee, statement, cardGrid, quotes, splitKi, codeFiles, splitMcp, capabilities, splitSprachen, coreScrolly, stackChips, fundament, teamAbout, faqBlock, cta, factsBand, pageHeader, featureList, compareTable, race, contactSplit, geoMap, eigenbau, legal, textContent, archive, mediaBlock, gallery, formBlock, newsletterBlock, snippetEmbed, testimonialsBlock, mediaText, cards, searchClient, html, styledHtmlBlock, datasetCitation. Use listAvailableBlocks for per-block field schemas.
Inputs: id, page, sort, depth, +10 more.
findPostsCInspect
Blog-style articles. content is Lexical richText (call listAvailableBlocks for richTextFeatures + inline-block list). PUBLIC AUTHOR: by default published AS the organization — set authorMode: "organization" (brand voice from the branding global). To publish AS a team member, set authorMode: "person" + teamAuthor: <team-member-id> (find the id via the team find tool — it lists members with id + name) and call getPersona(<id>) for their writing voice (style). The author field (relation to users) is the INTERNAL "created by" only — NOT the public author; it defaults to the user your key is bound to. Use categories+tags for taxonomy. SCHEDULING: to publish (or unpublish) at a FUTURE date, keep _status:"draft" and call the schedulePost tool with publishAt/unpublishAt — do NOT set a future publishedAt (the post stays a draft until the scheduled time, then auto-publishes). INCOMING LINKS while draft/scheduled: only link an anchor phrase that already exists in a published text (no new sentence, no rewording; new referring sentences only on publication day). posts collection. Drafts: yes (publish workflow; a find WITHOUT draft:true returns the published main-table state only). Public read: yes via /api/agents. MCP capabilities: find, create, update, delete. ⚠ delete is PERMANENT (hard-delete; the admin Trash does NOT catch MCP deletes). Write ONE document at a time: wait for the response of each create/update/delete call before sending the next — parallel writes can fail with a database deadlock (HTTP 500, nothing saved; retry that call alone). Before writing public text here, call getWritingGuidelines with this collection and the locale (brand voice, writing rules, fixed terms, SEO rules, taken keywords, field rules). Fields: title (localized), slug (localized), excerpt (textarea, localized), coverImage (upload: media|brand), content (richText, localized), author (rel: users), authorMode (enum: organization (Organisation)|person (Person (team member))), teamAuthor (rel: team), publishedAt (date), categories (rel-array: categories), tags (rel-array: tags), relatedFaqs (rel-array: faqs), robots (enum: index (Index (default))|noindex (noindex (hide))), searchText (textarea, localized, ≤60000 chars, redacted), meta (group: title, description, focusKeyword, secondaryKeyword1-4, searchIntent, eeatInsight, image — localized).
Inputs: id, page, sort, depth, +10 more.
findTagsAInspect
Flat keyword-tags for posts. Field is name (not title). Slug auto-generated if omitted. tags collection. Drafts: no. Public read: yes via /api/agents. MCP capabilities: find, create, update, delete. ⚠ delete is PERMANENT (hard-delete; the admin Trash does NOT catch MCP deletes). Write ONE document at a time: wait for the response of each create/update/delete call before sending the next — parallel writes can fail with a database deadlock (HTTP 500, nothing saved; retry that call alone). Before writing public text here, call getWritingGuidelines with this collection and the locale (brand voice, writing rules, fixed terms, SEO rules, taken keywords, field rules). Fields: name (localized, required), slug, usageCount (number).
Inputs: id, page, sort, depth, +10 more.
findTeamBInspect
Team members / people (Person profiles). A person gets a public author page /team/ once they have at least one published post (authorMode: "person" + teamAuthor); its URL is canonicalUrl in the response. slug is always generated from name (read-only, localized); there is deliberately NO index page /team (a team overview, if any, is a regular page). The author page shows photo, role, bio and the person's published posts automatically. name is required (language-neutral); role/department/bio/quote/qualifications are localized. socialLinks + languages are one-per-line textareas (→ Person sameAs / knowsLanguage). email/phone are public when filled. photo is an upload. Manual order via order (team block, not the author page). Site-specific profile fields: knowsAbout (see the field descriptions; shown on the author page and in its Person markup). Linked to the site Organization (worksFor) in JSON-LD; referenced as a post author — use the find tool to list members with their id (needed for a post's authorMode: "person" + teamAuthor + getPersona(<id>)), which makes the post byline link to the author page and its author point at the person's profile node. Never set a person as author without evidence they wrote the post (default: organization). style (the writing voice) is REDACTED in find responses — read the real value via getPersona(<id>). team collection. Drafts: yes (publish workflow; a find WITHOUT draft:true returns the published main-table state only). Public read: yes via /api/agents. MCP capabilities: find, create, update, delete. ⚠ delete is PERMANENT (hard-delete; the admin Trash does NOT catch MCP deletes). Write ONE document at a time: wait for the response of each create/update/delete call before sending the next — parallel writes can fail with a database deadlock (HTTP 500, nothing saved; retry that call alone). Before writing public text here, call getWritingGuidelines with this collection and the locale (brand voice, writing rules, fixed terms, SEO rules, taken keywords, field rules). Fields: name (required), slug (localized), role (localized), bio (richText, localized), quote (textarea, localized), photo (upload: media|brand), order (number), kind (enum: member (Team member (shown + author))|display (Display only (shown, not author))|author (Author only / guest (not shown)), required), department (localized), email, phone, socialLinks (textarea), qualifications (textarea, localized), languages (textarea), knowsAbout (textarea, localized), style (textarea, localized, redacted), publishedAt (date), robots (enum: index (Index (default))|noindex (noindex (hide))), searchText (textarea, localized, ≤60000 chars, redacted), meta (group: title, description, focusKeyword, secondaryKeyword1-4, searchIntent, eeatInsight, image — localized).
Inputs: id, page, sort, depth, +10 more.
findTestimonialsAInspect
Customer quotes for use in testimonials-block. author + quote required. Optional headline = one-line key statement shown above the quote — verbatim from the quote or approved by the customer, never invented. Rating 1-5. featured: true makes it prominent. rating = stars shown on THIS quote only; the site never derives an overall rating from quotes (no JSON-LD, no brand.json rating). Has drafts: use _status:"published" to go live (draft stays hidden); schedule a future date via the schedulePost tool (collection "testimonials"). testimonials collection. Drafts: yes (publish workflow; a find WITHOUT draft:true returns the published main-table state only). Public read: yes via /api/agents. MCP capabilities: find, create, update, delete. ⚠ delete is PERMANENT (hard-delete; the admin Trash does NOT catch MCP deletes). Write ONE document at a time: wait for the response of each create/update/delete call before sending the next — parallel writes can fail with a database deadlock (HTTP 500, nothing saved; retry that call alone). Before writing public text here, call getWritingGuidelines with this collection and the locale (brand voice, writing rules, fixed terms, SEO rules, taken keywords, field rules). Fields: author (required), role (localized), company, avatar (upload: media|brand), headline (localized), quote (textarea, localized, required), rating (number, 1–5), featured (boolean), publishedAt (date).
Inputs: id, page, sort, depth, +10 more.
getWritingGuidelinesAInspect
CALL THIS FIRST before you write or rewrite ANY public text (pages, team or any other collection with public text you have tools for) — once per task and target collection/locale. Returns the complete writing brief the site's own AI uses: brand voice and addressing, the house writing rules (internal — follow them, never state them), fixed terms that stay unchanged in every language, the terms & language rules of the requested locale, available placeholders, the author voice (with teamAuthor), the site-wide SEO rules (branding.aiSeoRules), the SEO meta rules (title/description length, title suffix), the keywords already taken in this locale (do not reuse) with their doc ids for internal links, the fields of the collection and how to write them via MCP. Pass topic (a few words or your draft) so the most related taken keywords come first. A long brief comes in parts: when the first line shows parts > 1, call again with the same arguments and part: 2, 3 … and read ALL parts before writing. The response is INTERNAL guidance — never publish it.
Inputs: collection (required), part, topic, locale, +1 more.
reportContentIssueAInspect
Report missing, wrong or outdated content on this website to the site operator. Not for security vulnerabilities. Use it when you notice a fact that is wrong, outdated or missing, a broken or misleading statement, or a gap a visitor would expect to be covered. A person reviews every report; nothing changes automatically. One clear report per issue — name the page (where) and what would be correct.
Inputs: message (required), kind, where, sender.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
Knowledge base MCP for AI agents on iknow.dev. Search, read, and maintain via OAuth.
Brivvy MCP (OAuth): on-brand AI content, voice rules, templates, glossary.
Web scraping, extraction and search for AI agents. Hosted MCP with workspace API keys.
Efficient, MCP-native headless CMS. Agents draft entries and roll back revisions over OAuth.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceConnects MCP-compatible AI clients to landing-page intelligence, evidence packs, improvement suggestions, experiment drafting and review, preview scans, instrumentation validation, site configuration, and integrations. Customer-facing writes require explicit OAuth scopes and confirmation.MIT

apexapi-mcpofficial
AlicenseAqualityDmaintenanceEnables calling 120+ AI models and reading live web pages (scrape, crawl, structured extract) from any MCP client using one API key.9143 npmMIT- AlicenseNot gradedqualityCmaintenanceEnables AI clients to access a marketing workspace’s memory, mission, training, prompts, skills, routines, research, saved media, and shop-building actions through an OAuth-secured MCP endpoint.1MIT

cob-rag-mcpofficial
AlicenseNot gradedqualityBmaintenanceEnables MCP clients such as Claude, Cursor, VS Code, and n8n to search and ask questions against a business's own self-hosted knowledge base, returning cited passages and answers while scoping access by API key.MIT
Glama MCP Gateway
Add one secure layer between your agents and this server.