- Starting price
- Free public endpoint
- Free tier
- Yes
- Platforms
- Remote MCP server · Claude-compatible clients · Cursor · developer tools
- Developer
- Cameron Pak / faith.tools
- Launched
- 2025
- Updated
- Aug 9, 2026
The verdict
Bible MCP does one narrow job well: an AI client supplies a book, chapter, verse range and translation ID, and the server retrieves real text from AO Lab’s Free Use Bible API instead of asking the language model to reproduce Scripture from memory. It is free, simple and safer than hallucinated quotation. Production currently exposes four tools, not the three described in older material, and plain verse results omit the reference, translation and license metadata a downstream agent most needs. Use it with explicit attribution, validation, caching and fallbacks—not as a complete Bible platform or proof that the model chose the right passage.
Try Bible MCP ↗Opens bible-mcp.faith.tools
Bible MCP is a public remote Model Context Protocol server created by Cameron Pak, the founder of faith.tools. An MCP-capable assistant or developer client connects to `https://bible-mcp.faith.tools/mcp`, discovers a small tool schema and can request Scripture text from the Free Use Bible API operated by AO Lab. The purpose is precise: language models are good at recognizing the idea of a familiar verse but can merge translations, drop words or invent a reference. A tool call moves exact text retrieval out of model memory and into a structured data source.
The live server identified itself as `bible-mcp` version 0.1.5 and negotiated MCP protocol version 2024-11-05 during research. Its current `tools/list` result exposed four tools: `get_available_translations`, `get_verse`, `get_verse_with_ui` and `get_chapter`. The public project README says the server provides “two tools” and then lists three, while the faith.tools description also says three. The additional UI-capable verse tool and the production schemas show that the deployment has moved beyond those descriptions. The working product should be treated as the live discovery response, not a static directory count.
The lookup schema accepts one of 66 Protestant-canon book names, a positive chapter, a starting verse, an optional ending verse and a translation ID. Berean Standard Bible is the default. Translation discovery takes a three-letter language tag; the English call returned 51 current records, including complete Bibles, testaments, portions, Septuagint editions and editions containing a deuterocanon. Each translation record includes its identifier, names, website and license URL. The wider upstream API supports many languages, so “hundreds of translations” describes the changing AO Lab catalog rather than a fixed promise bundled inside Bible MCP.
A direct John 3:16 BSB call returned the verse body promptly, but the plain result contained only text—not “John 3:16,” “BSB,” the translation’s license URL, publisher or upstream provenance. An impossible John 999 request passed the broad numeric input schema and returned the unhelpful text “Error: Unknown error.” The book field is tightly enumerated, but chapter and verse maxima are generic large integers rather than book-specific bounds. The tool therefore improves wording accuracy while leaving citation completeness, reference selection, validation and error recovery to the client.
The public Val Town project makes its files and history visible and allows platform remixing, but its displayed eight-file tree contains no LICENSE file. Publicly readable source is not automatically open-source permission for reuse outside whatever Val Town’s platform terms allow. There is also visible deployment drift: the Val Town README describes a Deno/Hono/Val Town stack and three tools, while live response headers identified Cloudflare in front of a Render origin and the endpoint returned four tools. Our 4.2 rating is a researched editorial assessment based on public code, protocol discovery, sample calls, upstream documentation and registry data—not a load test, source-to-production audit, licensing opinion or guarantee for any MCP host.
✓ The good
- Retrieves real text - the model asks a Scripture data service for wording instead of reconstructing a verse from probabilistic memory
- Very small tool surface - translation discovery, verse range, chapter and a UI verse response are easy for clients to understand
- No key or account at the direct endpoint - a developer can connect without provisioning a Bible MCP credential
- License discovery exists - translation results include source websites and license URLs rather than flattening every edition into “free”
- Structured input - book names, positive integers and translation identifiers reduce many malformed-reference failures before the upstream request
- Broad upstream catalog - language-filtered AO Lab data reaches well beyond the default Berean Standard Bible
- Source is inspectable - implementation files and version history can be read and remixed on Val Town
- Low prompt-injection exposure - tools accept constrained Scripture coordinates, not arbitrary web pages or user-authenticated accounts
✗ Watch out
- Plain verse output loses citation metadata - the response body does not identify reference, translation, license or source
- Documentation is stale and self-contradictory - it says two tools, lists three, while production exposes four
- Source and deployment do not fully match - public files describe Val Town while live headers reveal a Cloudflare and Render path
- No explicit code license in the project tree - visible or remixable source should not be called unambiguously open source
- Validation is shallow - impossible chapter and verse coordinates can reach the upstream call and collapse into “Unknown error”
- The model can still choose the wrong passage - exact retrieval does not verify that the agent interpreted the user’s question correctly
- No product privacy or service notice - logging, retention, subprocessors, uptime, incident handling and endpoint change policy are undocumented
- Not a search or study API - no keyword search, cross-references, notes, morphology, commentary or semantic answer checking
Best for
- Developers adding exact Scripture retrieval to an MCP-capable assistant or coding workflow
- Low-risk prototypes that need public or permissively usable Bible editions without managing another API credential
- Agents that already know a passage reference and need to quote or display it accurately
- Teams prepared to preserve translation attribution, validate coordinates and operate their own cache and fallback
- Educators demonstrating the difference between model recall and tool-grounded retrieval
Avoid if
- You need a contractual Bible API SLA, support channel, DPA, fixed quotas or enterprise security evidence
- You need commercial translations whose publishers do not expose them through the upstream free-use catalog
- You need keyword search, topic discovery, semantic retrieval, commentary, cross-references or original-language data
- You cannot implement reference validation, attribution, caching, monitoring and a no-tool fallback
- You are an ordinary reader looking for a Bible interface rather than developer infrastructure
What Bible MCP is
MCP lets a compatible AI host discover named tools, read their JSON schemas and call them during a conversation or agent workflow. Bible MCP implements that interface for Scripture coordinates. It does not generate an interpretation. It translates a model-produced function call into a Free Use Bible API request and returns the resulting text or structured translation inventory.
The distinction between server and host matters. Bible MCP does not control whether Claude, Cursor or another client calls the tool, displays the translation label, rewrites the returned verse, places it in a harmful answer or sends conversation data to its own provider. It also depends on AO Lab for the Bible catalog and on its current hosting chain for availability. Evaluate all three layers: AI host, MCP server and upstream Scripture source.
Why exact retrieval solves one hallucination problem—and not the question before it
Without a tool, a model may answer “John 3:16 says…” from statistical memory. Familiar wording can be a blend of KJV, NIV and a paraphrase, and obscure verses are more likely to be invented. Bible MCP constrains the action to book, chapter, verse and edition, then requests text from a maintained corpus. That makes the returned string traceable to a real translation and avoids using copyrighted model training as a substitute for licensed distribution.
The model still decides that John 3:16 is relevant, constructs the coordinates and explains what the text means. It can select John 3:17 by mistake, use the wrong translation ID, ignore a result, truncate it or wrap exact wording in a false claim. A trustworthy client should show the requested reference and edition beside unmodified text, preserve the license link, distinguish quotation from commentary and ask the user to confirm high-consequence references. Tool grounding narrows the failure surface; it does not confer biblical judgment on the agent.
Verse and chapter tools: narrowly useful output needs a citation envelope
`get_verse` accepts a book, chapter, first verse, optional last verse and translation. `get_chapter` fetches a complete chapter. `get_verse_with_ui` exposes a display-oriented response for hosts that support MCP application interfaces. The 66-book enum prevents arbitrary book strings, and positive integer requirements eliminate zero or negative coordinates. The default BSB gives a client a useful result even when it does not deliberately choose an edition.
The plain verse call we inspected returned only the verse sentence. A client cannot safely infer the response’s citation envelope after the fact: it should receive canonical reference, requested and resolved translation IDs, publisher, license URL, upstream revision and perhaps verse boundaries in structured content. Chapter and verse ranges should be validated against the chosen edition, with machine-readable error codes such as `chapter_out_of_range` instead of “Unknown error.” Those changes would make the tool safer for autonomous composition.
Translation discovery: live license links are better than a hard-coded name list
`get_available_translations` filters the upstream catalog by an ISO 639 three-letter language tag. The English request returned 51 records during review. Each included an ID, display names, a source website, a license URL and short name. The selection mixes complete Bibles, New Testaments, portions, historic editions, contemporary open editions, Septuagint translations and editions with additional books, so “available” does not mean equal coverage or identical canon.
Clients should query rather than assume, then inspect the chosen edition’s license and book coverage. A translation that can be fetched freely may still require attribution, restrict modification or differ by jurisdiction. Bible MCP’s own book enum stops at the 66-book Protestant list even when an upstream edition advertises a deuterocanon, so catalog membership exceeds tool-addressable canon. Cache license metadata with every text record and never present the upstream collection as public domain as a whole.
Remote Streamable HTTP: easy connection, concentrated operational dependency
A modern MCP host can connect to one HTTPS endpoint rather than installing a local process, runtime or secret. The server answered discovery and calls without authentication and allowed cross-origin requests. Smithery reports usage by clients such as Cursor, Claude Code, Claude.ai and its own playground, and its registry supplies another installation route. For a hackathon or personal assistant, that is excellent friction reduction.
Remote simplicity means every call leaves the client and depends on DNS, Cloudflare, the origin host, the MCP application and AO Lab. There is no published status page, SLA, retention schedule, incident policy, versioned endpoint or deprecation window. The response headers and public source also point to different hosting stories. Production applications should set timeouts, cap retries, validate content, cache permitted text, monitor changes to `tools/list` and retain a direct-upstream or local-corpus fallback.
Pricing
Direct Bible MCP endpoint
Free
The public `/mcp` endpoint required no API key or Bible MCP account during research. Its response advertised a 5,000-request rate-limit window, but no public page defines the window length, fair-use policy, guaranteed quota, commercial entitlement, support, uptime, version-retention period or remedies. A header observed today is operational evidence, not a service contract.
Smithery connection
Registry/platform terms may apply
Smithery lists the server and can broker installation or managed connections for supported clients. Bible MCP itself remains free, but an account, managed connection or related Smithery service follows Smithery’s current plans and policies. The direct endpoint avoids that intermediary when the client supports remote Streamable HTTP configuration.
Self-managed adaptation
Hosting plus legal and engineering work
The implementation is visible and Val Town offers a Remix workflow, but the project tree does not show an explicit code license. Obtain permission or clarification before treating it as generally reusable open-source software. A self-host also inherits AO Lab translation licenses, infrastructure, monitoring, abuse control, updates and attribution work.
Bible MCP does not charge for calls and exposes no checkout, subscription or donation tier. For a developer, that removes procurement and secret-management friction. It does not make capacity unlimited or guaranteed. The observed limit header needs a published definition before it can be used in architecture calculations.
The upstream Free Use Bible API is built around text that can be distributed under stated permissions, but “free use” is a licensing strategy, not a declaration that every record is ownerless. Translation results deliberately include license URLs. Engineering time must be budgeted to surface attribution, honor version-specific terms and prevent a downstream user from exporting more than an edition permits.
A Smithery-managed route may add account or platform economics that do not belong to Bible MCP. Compare direct endpoint support in the intended host before adding a registry intermediary. If an organization uses Smithery Connect or another gateway, that vendor becomes part of the privacy, availability and credential story.
Self-hosting is not automatically authorized by the word “public.” The Val Town project can be viewed and remixed on that platform, but an explicit OSI-style code license was not present in the file tree we reviewed. Ask the builder to add MIT, Apache-2.0 or another intended license and clarify whether production code is the same revision before forking outside Val Town.
The largest hidden cost is reliability. An assistant that promises exact Scripture but silently falls back to model memory when the endpoint fails is worse than one that discloses uncertainty. Implement a hard rule: when the tool cannot return validated text, provide the reference without quotation or direct the user to a trusted Bible. Accuracy includes knowing when no source was reached.
Where Bible MCP falls behind
Documentation accuracy is the first repair. The README says “two tools,” displays three bullets and production reports four. The landing page contains only a code link, endpoint and thank-you. Publish an exact quick start for major hosts, current tool table, server version, protocol versions, examples, error behavior and a dated change log.
The response should be citation-first. Plain `get_verse` output needs structured fields for canonical reference, verse numbers, selected translation, full translation name, license URL, source URL, upstream version and unchanged text. An agent should not reconstruct those facts from its own call arguments or a separate lookup.
Input validation needs canonical awareness. The schema permits any positive chapter or verse up to JavaScript’s safe integer maximum and validates only the book name against a 66-item list. Impossible coordinates should fail before an upstream request with a stable error code, valid range and suggestion. Common aliases such as Psalm/Psalms and Song of Songs also need documented normalization behavior.
“Open source” needs a license. The code is public on Val Town and the platform has a Remix action, but the eight visible files include no LICENSE. The publisher should state the code license, upstream data boundary, icon license and contribution terms so source availability and reuse permission are not conflated.
Deployment provenance is unclear. Public documentation says Deno and Hono on Val Town; the live endpoint exposed Cloudflare and a Render origin, a different server version and a fourth tool. Publish the production repository or signed commit, build process, hosting regions, dependency inventory and release mapping.
Privacy and operations have no product notice. Even constrained calls expose IP address, client headers, timestamps, selected references and translation to infrastructure. State what Bible MCP, Cloudflare, the origin and AO Lab log, how long records persist, whether data is used for analytics or abuse protection and how a user can report a vulnerability or incident.
There is no service commitment. A free personal endpoint can reasonably be best effort, but clients need a status page, health endpoint, published rate-limit window, maintenance ownership, last update and deprecation policy. Registry uptime is useful external evidence, not a promise from the operator.
The tool set is intentionally incomplete. It cannot search for a phrase, resolve a natural-language allusion, return cross-references, compare editions, supply paragraph boundaries, explain a license, expose morphology or assess interpretation. Clients should not send those jobs through a verse-coordinate tool and pretend the response answers them.
Bible MCP vs. Free Use Bible API vs. API.Bible and Bible Brain
The Free Use Bible API is the direct upstream and offers ordinary HTTP endpoints, downloads, commentaries and datasets beyond Bible MCP’s four tools. A conventional application should usually integrate it directly: fewer intermediaries, full metadata and more control. Bible MCP wins when the consumer is specifically an MCP host and a tiny discoverable tool surface is more valuable than API breadth.
API.Bible from American Bible Society provides a managed developer API and access to many licensed Bible editions under its own key, agreements and limits. It is the stronger choice when an organization needs documented credentials, catalog controls and a conventional support relationship. Bible MCP is easier for a free prototype and is restricted to what the open upstream can legally serve.
Bible Brain API from Faith Comes By Hearing focuses on a vast multilingual audio, text and video corpus, especially for global Scripture engagement. It requires developer integration and attention to permissions, and its media workflow is far richer. Bible MCP does not return audio or video; its advantage is a four-tool language-model connection, not global media coverage.
YouVersion Platform and other newer Scripture services add SDKs, embeddable readers, licensed editions and organizational partnerships. They carry more product surface and governance. Bible MCP is best understood as a small adapter that proves a good pattern: let the model call a Scripture source. A mature team may copy the pattern while choosing an upstream and hosting arrangement that matches its obligations.
The bottom line
Bible MCP is a good idea implemented at the right initial size. It turns “quote this verse” into a constrained call against a real Bible corpus and keeps API credentials out of a first experiment. Its translation-discovery tool preserves license links, and the public endpoint worked without setup. The product’s own precision standard now needs to extend beyond verse wording. Current documentation cannot agree on the tool count; production has moved beyond the visible Val Town description; the code tree lacks an explicit license; errors are generic; and plain verse output strips the exact citation and license context an AI answer needs. Connect it for prototypes and personal workflows, but wrap it carefully. Pin or monitor the discovered schema, validate book-specific coordinates, show reference and translation beside unmodified text, preserve upstream license metadata, never fall back silently to model memory, and provide a direct source link. For production, ask for a privacy notice, status and security contact, explicit code license, versioned deployment, rate-limit definition and structured errors. The safest lesson from Bible MCP is broader than this server: a model should retrieve Scripture rather than impersonate a Bible, and the retrieval result should carry enough provenance that a human can verify it.
Alternatives to Bible MCP
Free Use Bible APIBible MCP’s direct upstream, with conventional HTTP access, downloads and broader Bible datasets for developers.
API.BibleA managed American Bible Society developer platform for licensed Scripture text and production integrations.
Bible Brain APIA global text, audio and video Scripture API with much broader multilingual media coverage.YouVersion Platform ServicesAPIs, SDKs and an embeddable reader backed by a large licensed Bible catalog and ministry platform.route.bibleAn app-agnostic linking standard for sending readers to a passage in their preferred Bible experience.Frequently asked questions
What is Bible MCP?
Bible MCP is a free remote Model Context Protocol server created by Cameron Pak. An MCP-capable AI host can call it with a Bible book, chapter, verse range and translation ID. The server retrieves real text from AO Lab’s Free Use Bible API so the model does not have to reproduce Scripture from memory.
How many tools does Bible MCP provide?
The live server exposed four during research: `get_available_translations`, `get_verse`, `get_verse_with_ui` and `get_chapter`. Older public descriptions say three, and the README says “two” immediately before three bullets. MCP clients should trust current tool discovery and monitor it for changes rather than hard-code the directory count.
Is Bible MCP free?
The direct endpoint required no account or API key and has no published paid tier. A response header showed a 5,000-request rate limit, but the operator does not define its window or guarantee capacity. Smithery or an AI host may have separate account and pricing terms. Translation licenses still apply even when the server call costs nothing.
Which Bible translations does Bible MCP support?
It queries the changing Free Use Bible API catalog by three-letter language code and defaults to the Berean Standard Bible. An English discovery call returned 51 records, including complete Bibles, portions and editions with different canons. Query the live catalog, inspect each license and coverage, and do not assume every upstream book is addressable through the server’s 66-book enum.
Does Bible MCP guarantee accurate Bible quotations?
It materially improves wording accuracy by retrieving from a real translation source. It does not guarantee that the AI selected the right reference, passed the desired edition, preserved the result or interpreted it correctly. Plain verse output also omitted reference and translation metadata in our check. Display and verify the passage context yourself.
Is Bible MCP open source?
Its implementation files and history are publicly visible on Val Town and the platform offers a Remix action. However, the visible project tree did not contain an explicit LICENSE file. Source-visible and platform-remixable are not automatically the same as broad open-source reuse permission. The publisher should add a clear code license and production revision.
Is Bible MCP private and reliable enough for production?
Inputs are narrowly limited to Scripture coordinates and translation IDs, so they are less sensitive than free-form conversations. The public product still has no specific privacy, retention, subprocessor, SLA, status, security or deprecation notice. Use timeouts, structured validation, permitted caching, monitoring and a fallback; never send the user’s whole pastoral conversation as an unnecessary tool argument.
Sources & further reading
More Bible Study Software
VerseLink BibleVerseLink Bible keeps Scripture, sermon notes, highlights and original-language helps entirely on your device with no account, tracking, ads or subscription—but local-first privacy makes manual backup your responsibility.
Ezra Bible AppA serious free, open-source study workspace built around topical verse collections, notes and the SWORD module ecosystem, now spanning desktop and mobile—but resource licensing varies, sync depends on Dropbox, and its 2022 privacy template contradicts the current App Store disclosure.
Rava: Discover God’s StoryA free, multilingual facilitator that lets almost anyone press play and guide a Discovery Bible Study with little preparation—excellent for reproducible group participation, but not a replacement for wise leadership, contextual interpretation or a modern privacy policy.
ARMA CoursesARMA makes serious Bible and theology teaching approachable through short, energetic courses—valuable for guided learners, but narrower and less transparent than a seminary library.
Storyverse InstituteStoryverse makes Bible, theology, soul care and ministry training approachable through 200+ videos, community and cohorts—but its age, collection and deletion disclosures conflict too sharply for informed enrollment.
The Word One to OneA polished, completely free guide for reading John and Acts with a friend—excellent for a trusted conversation, but narrow in scope and incomplete on app privacy and safeguarding.