- Starting price
- Free
- Free tier
- Yes
- Platforms
- Web · Open-source code
- Developer
- Jon Boldt
- Launched
- 2023
- Updated
- Aug 9, 2026
The verdict
Biblos makes semantic search understandable and genuinely useful: describe an idea, receive Bible sections ranked by similarity, and open the surrounding chapter. Its current BGE service found “love your enemies” immediately, while also returning thematically adjacent but potentially misleading passages—exactly why similarity cannot replace exegesis. The expanded daily reader is generous and local-first for progress, but AI summaries have opaque providers and the site offers no privacy or terms page. Use it to discover passages and sustain a plan, then verify context elsewhere.
Try Biblos ↗Opens biblos.app
Biblos is a free web application from developer Jon Boldt that helps readers find passages by meaning rather than exact wording. A query such as “hope when everything feels lost” is converted into a numerical embedding and compared with embedded sections of the Bible. Results are ranked by similarity and displayed with their references and text. This is useful when someone remembers a concept but not a phrase, uses different vocabulary from a translation or wants to explore thematic connections that a conventional word search would miss.
The current live site publishes unusually concrete model details. It uses BAAI’s BGE-large-en-v1.5 embedding model, described as 335 million parameters with 1,024-dimensional vectors. Its hosted health endpoint reported that the model, tokenizer and embeddings were loaded and that all 66 books were available during review. The browser requests up to 50 candidates, filters scores below a current threshold and initially shows a small set, with controls to reveal more. Results are section or chapter-chunk passages, not guaranteed single-verse answers.
A live query for “love your enemies” ranked Matthew 5 first at about 0.71 similarity, followed by passages including Romans 12 and Luke 6—strong matches. It also surfaced Psalm 21’s language about enemies and James 4, which are related at the word or topic level but can point in very different interpretive directions. That is an excellent illustration of both the tool’s power and its boundary. A vector score estimates textual-semantic closeness; it does not determine speaker, covenant setting, genre, command, promise, irony, fulfillment or how passages relate theologically.
Biblos has grown beyond its older public repository description. The live Next.js site includes direct reference lookup and a daily-reading area with a 365-day chronological plan, a classic five-day-per-week plan, local progress, ESV audio links and selections from early Christian writers. Optional AI buttons summarize search results or daily commentary. The main search page also includes Vercel Analytics. We found no live privacy, terms or about page, and the current summary provider and retention are not documented. The GitHub repository remains useful context but describes an older Streamlit, Chroma, instructor-large and Claude architecture; it should not be mistaken for exact documentation of the current deployment.
✓ The good
- Semantic search solves a real problem - concepts can find passages even when the query does not match exact Bible wording
- Model details are visible - BGE-large-en-v1.5, dimensionality and parameter count are stated on the live product
- No account is required - search and daily reading can begin without registration
- Results preserve biblical text - users see sections and references rather than an AI answer with no evidence
- Direct references are supported - known citations such as John 3:16 can bypass semantic search
- Daily plans add long-term value - chronological and five-day rhythms turn a search demo into a repeat-use reader
- Progress stays in the browser - plan, day and completion choices are stored locally rather than tied to a profile
- Source code is public - the repository provides inspectable history and a useful foundation for technical readers
✗ Watch out
- Similarity can mislead - enemy-language passages can rank near “love your enemies” despite serving a different purpose
- Results are chunks, not interpretations - a high score does not identify the exact relevant verse or explain context
- AI summary provenance is opaque - the current provider, prompt, sources, retention and safety process are not disclosed
- No privacy policy found - analytics and server requests are not mapped to a public data-handling statement
- No terms page found - availability, liability, user rights and service-change rules are unclear
- Live code and repository diverge - the public README documents an older architecture, which can confuse evaluators
- The repository license is noncommercial - CC BY-NC 4.0 is source-available but not a permissive license for business reuse
- No account sync - local progress is private and simple but can disappear with cleared browser storage or a new device
Best for
- Readers who remember a biblical idea but not its exact wording
- Teachers brainstorming passages to investigate before preparing a lesson
- Developers learning how vector search can support Bible discovery
- People who want a free chronological or five-day reading plan without an account
- Users willing to inspect every result in its full literary context
Avoid if
- You need authoritative interpretation, counseling or doctrinal conclusions from a similarity score
- Your research requires multiple translations, Greek or Hebrew, manuscript data and named commentaries
- You require synchronized reading progress across devices
- Your organization needs a privacy agreement, SLA or documented AI subprocessor before use
- You want to reuse the code commercially under a permissive open-source license
What Biblos is
Biblos is a retrieval tool. The embedding model transforms a query and Bible text chunks into vectors, then measures which passages occupy nearby positions in that learned semantic space. This allows “feeling abandoned by God” to retrieve relevant language even if the exact query phrase never appears in the translation. The current corpus is the 66-book World English Bible Protestant edition, identifiable both in live wording and the source input data.
The daily section is a separate reading application built around structured plans. It computes the current plan location, loads Bible chapters, records completion in localStorage and can request audio or a concise summary. It also includes readings associated with Church Fathers. These functions make Biblos a hybrid: part semantic concordance, part browser-local Bible plan. Neither part is a substitute for licensed pastoral care or a scholarly research environment.
Why concept-first Bible discovery is valuable
Traditional search works beautifully when a reader remembers “still small voice” or knows to search for “justification.” It fails when the remembered idea uses different words, the translation changed the phrasing or the reader cannot name the relevant theological term. Biblos accepts ordinary language and retrieves candidate passages, lowering the vocabulary barrier between a question and the biblical text.
Its best design decision is showing the text instead of replacing it with a synthetic answer. The reader can see why a result might match, reject a false thematic neighbor and open the chapter. Biblos should reinforce that discipline by highlighting the most influential sentences, showing score explanations, diversifying results across books and adding a prominent reminder that relevance ranking is discovery—not authority, interpretation or divine guidance.
BGE semantic search: fast thematic discovery with measurable noise
The live backend identifies BGE-large-en-v1.5 with 1,024-dimensional embeddings and reports all 66 books loaded. The front end requests a larger candidate set, removes low-scoring items and progressively reveals results. A direct-reference parser handles queries that already contain a citation, avoiding needless semantic inference. In our sample, the desired Sermon on the Mount passage ranked first and closely related New Testament texts appeared near it.
The same sample produced results whose use of “enemy” moved in another direction. This is not necessarily a model defect; vector retrieval is doing association, not biblical theology. Product safeguards should include book-and-testament filters, lexical plus semantic hybrid search, duplicate suppression, visible thresholds, feedback, full-context previews and evaluations across common pastoral queries. A “why this matched” view would teach users how to judge rather than merely trust the rank.
Daily reading plans: more complete than the old repository suggests
The current live code exposes two plan structures: a 365-day chronological sequence based on Blue Letter Bible’s plan and a classic five-day-per-week schedule. The chronological data includes validation intended to cover the complete set of chapters exactly. Users can choose the plan, move between days or weeks, use an automatic-today setting and mark progress. Selections remain in browser storage without an account.
Local-first state is an excellent privacy default but a fragile backup strategy. Clearing site data, changing browsers or losing a device can erase the record. Biblos should offer an explicit JSON export and import, explain the calendar anchor and missed-day behavior, credit every source beside the plan, and make audio availability and translation differences clear. Optional encrypted sync could remain opt-in rather than turning all readers into accounts.
Audio, Church Fathers and AI summaries: rich additions with mixed provenance
The daily reader can request ESV audio and show selections from early Christian writers, with a user choosing among available figures. Optional summary controls send top search results or daily commentary material to server endpoints for condensation. A summary may help a reader orient to a difficult historical passage, especially when the original style is unfamiliar.
These layers need clearer sourcing. The live application should name the edition and translator of every patristic text, link the exact work and section, explain ESV API and attribution requirements, and identify the current AI provider. Before a summary request, it should preview what text is transmitted, state retention and offer a local “do not use AI” default. Condensation can omit qualifications or make a historical writer sound more uniform than the source.
Pricing
Semantic Bible search
Free
The public site accepts natural-language queries, ranks World English Bible sections through its hosted BGE embedding service and opens surrounding chapters. No account, subscription or published query allowance was required during this review.
Daily reading tools
Free
Chronological and classic five-day plans, browser-local progress, ESV audio requests, patristic readings and optional summaries are included in the live site. Progress is device-local, so users should not expect cloud backup or cross-device continuity.
Source reuse
Free for license-compliant use
The public repository is available under Creative Commons Attribution-NonCommercial 4.0. That permits defined sharing and adaptation with attribution for noncommercial purposes; it is not blanket authorization to incorporate the work into a commercial product.
Biblos is free to use, with no visible consumer account or premium plan. Semantic queries still consume shared hosted compute, so users should expect reasonable limits, occasional cold starts or outages even though no quota table is published.
The search backend runs through a public Hugging Face Space. Hosted model demonstrations can sleep, restart or change capacity. A free endpoint is a generous public service, not a service-level guarantee for an application that depends on every request completing at a fixed latency.
The daily reader adds no subscription, and browser-local progress avoids the financial and privacy cost of an account. Its tradeoff is portability. A user should treat current progress as convenient local state until export exists, not as a durable record guaranteed by the developer.
The source repository’s CC BY-NC 4.0 license matters to developers. It permits sharing and adapting under its terms for noncommercial purposes with attribution. Calling the project “open source” conversationally can obscure that the Open Source Definition generally does not permit a commercial-use restriction; “source available under a noncommercial Creative Commons license” is more precise.
Optional summaries may create the largest ongoing cost even when no price appears. Biblos could keep the service sustainable by documenting a rate limit, offering bring-your-own-provider keys or making deterministic retrieval the permanent free core. It should never quietly make a paid AI request necessary to read the Bible or keep plan progress.
Where Biblos falls behind
Publish privacy and terms pages. Identify analytics, search logs, IP handling, referrers, summary prompts, providers, retention, deletion and user rights, alongside availability and liability boundaries.
Document the live architecture separately from the repository history. The README still describes Streamlit, Chroma, instructor-large and Claude, while the current site exposes Next.js, BGE and a different hosted backend. Date and version both.
Add a transparent search evaluation. Measure known-reference recall, thematic relevance, contradictory near-neighbors, canonical coverage and performance across plain-language queries, then publish examples and failure cases.
Expose summary provenance and controls. Name the model and provider, preview transmitted inputs, prohibit training where applicable, state retention, cite the source text and let users disable all generative requests permanently.
Improve result context. Highlight the likely matching verse inside each chunk, display surrounding paragraphs on demand, diversify books, add translation labels and make similarity scores understandable without implying confidence in interpretation.
Protect daily progress. Add offline-capable export and import, backup warnings, plan-source attribution and an optional end-to-end encrypted sync path that does not become mandatory.
Clarify licensing language. State plainly that the repository is CC BY-NC 4.0 source-available software, distinguish code from Bible text and third-party content licenses, and identify what would require separate permission for deployment.
Biblos vs. Bible Vector Search vs. Blue Letter Bible
Bible Vector Search pursues the same core task: find Scripture through meaning rather than exact words. Interface, corpus, ranking, filters and transparency should decide between them. Biblos is especially appealing because it names the current embedding model and combines search with daily plans. Using both can be productive when a difficult query needs independent candidate lists rather than faith in one ranking.
Blue Letter Bible is a mature study site with conventional search, many translations, lexicons, interlinear tools, cross-references, commentaries, audio and reading plans. It is much stronger after a passage has been found. Biblos is faster when the starting point is an imprecise idea. A good workflow discovers in Biblos and investigates in Blue Letter Bible.
STEP Bible is a free scholarly reader designed around translation and original-language exploration. It does not need a generative summary to expose lexical and textual relationships. Its interface is denser and can require learning. Biblos wins immediate concept search; STEP wins depth, provenance and tools for testing what a passage actually says.
A conventional concordance or exact search remains essential. Semantic systems can retrieve paraphrases but also flatten distinctions. Exact search proves where a word or phrase occurs and supports reproducible counts. The strongest study practice alternates both: use embeddings to generate candidates, exact and original-language tools to test patterns, and full literary context to interpret them.
The bottom line
Biblos is a valuable free instrument precisely when its output is treated as a list of leads. The current model details are refreshingly specific, the live backend covers all 66 World English Bible books, and our sample query showed excellent first results alongside instructive semantic noise. That combination teaches a healthy lesson: vector search can overcome vocabulary mismatch, but it cannot tell a reader what Scripture means. The newer daily experience makes the site much more than a technical demo. Two serious plan structures, local progress, audio and historical readings could sustain ordinary use for a year. Before we recommend the complete stack without qualification, Biblos needs a privacy policy, current architecture documentation, clear AI-summary provenance and portable progress. We recommend it strongly for passage discovery, especially when paired with a conventional study Bible or original-language resource. Keep AI summaries optional, inspect the full chapter, notice who is speaking and why, and never convert a similarity score into a command, promise or theological conclusion by itself.
Alternatives to Biblos
Bible Vector SearchA dedicated meaning-based Scripture search alternative worth comparing when a query needs independent semantic results.
Blue Letter BibleA mature free reader with exact search, original languages, cross-references and commentaries for investigating Biblos results.STEP BibleA free scholarly Bible environment with strong language and translation tools, best after a relevant passage has been identified.
Bible GatewayA broad multi-translation reader with dependable conventional search, audio and plans but no equivalent semantic-first workflow.
LogosA professional research platform for users who need cited books, advanced search and original-language analysis beyond free discovery.Frequently asked questions
What is Biblos?
Biblos is a free Bible web app by Jon Boldt. It uses semantic embeddings to find passages related to a natural-language idea and also provides direct reference lookup, daily reading plans, audio and optional AI summaries.
Which Bible translation does Biblos search?
The current corpus is the 66-book World English Bible Protestant edition. The source data and live wording support that identification. ESV audio in the daily section is a separate feature and does not mean semantic results use ESV text.
Which AI model powers Biblos semantic search?
The live site identifies BGE-large-en-v1.5, a 335-million-parameter embedding model producing 1,024-dimensional vectors. Older repository documentation describes a different model and architecture, so use the live disclosure for the current service.
Are Biblos search results accurate?
They can be very useful but are relevance candidates, not authoritative answers. Our “love your enemies” test ranked Matthew 5 first and found other strong texts, while also surfacing enemy passages with a different direction. Always open and interpret the context.
Does Biblos require an account?
No. Search and daily reading work without registration. Reading-plan choices and progress are stored in the browser, which improves privacy but means they may not follow you to another device and can disappear if local data is cleared.
Is Biblos open source?
Its source repository is public under Creative Commons Attribution-NonCommercial 4.0. The code can be inspected and reused within that license, but the commercial-use restriction means it is more precise to call it source available than permissively open source.
Does Biblos have a privacy policy?
We did not find a live public privacy or terms page during this review. The search service receives query requests, optional summaries send content to server endpoints, and the main page includes Vercel Analytics, so the missing disclosure is material.
Sources & further reading
More AI Bible & Faith Tools
Magisterium AIA chatbot that refuses to answer Catholic questions from the open internet - and instead pulls every claim from 25,000+ Magisterial documents with citations.
AnorA white-glove intelligence layer that connects a church or nonprofit’s existing email, giving, event, and commerce systems, resolves constituent identities, and answers cross-system questions—powerful for organizations with fragmented data, but costly and governance-heavy because the “complete picture” is sensitive by design.
AskMyChurchA church-trained AI assistant with unusually thoughtful citations, anonymous questions, pastoral boundaries and crisis routing—stronger than a generic chatbot, yet paired with absolute accuracy claims and incomplete retention, vendor and compliance detail.
Bible BotA carefully instructed Christian companion inside ChatGPT, with unusually candid limits and pastoral safety rules—but still a probabilistic AI whose model, privacy, memory, and access belong to OpenAI’s platform.Faith.toolsA free, browsable directory of Christian apps, websites, and software - including the new wave of AI tools - built to help you discover resources for Bible study, prayer, and ministry.BibleAIA conversational AI Bible study tool that answers questions and surfaces relevant Scripture on demand - fast and easy to use, and best treated as a study aid you check against the text.