✦ Feason

Feason Context · Model Context Protocol

Christian context
your AI can inspect.

Feason is a context layer for AI agents—not another chatbot. It gives developer tools structured access to source-grounded Scripture and theological research, verifiable citations, named tradition perspectives, and permissioned study memory.

Start in five minutesUse the REST API
Remote MCP endpointhttps://www.feason.com/api/mcp

Detailed overview

One MCP server.
Three clear layers.

Feason separates shared evidence, personal context, and durable research so an agent can use each without confusing one for another.
01

Canonical context

Search a versioned corpus of Scripture, creeds, councils, patristic writing, historical theology, and confessional sources. Results preserve source, passage, edition, translation, license, URL, and verification metadata.

02

Private study context

With explicit permission, an agent can retrieve a user’s saved passages, notes, reading history, study progress, lecture notes, plans, and authored reflections. Personal data stays separate from canonical ranking.

03

Research workspace

Create private projects, save immutable evidence packets, curate exact passages, compare packet history, export citations, fork an inquiry, and create revocable read-only shares.

How it works

From tool call
to traceable evidence.

The model can reason over the response, but Feason controls retrieval, permissions, evidence boundaries, and provenance.
  1. 01

    The agent chooses a named tool

    Instead of asking a model to recall theology from training, the client calls a specific Feason tool such as search_sources, compare_traditions, or build_evidence_packet.

  2. 02

    Feason enforces the boundary

    OAuth and client credentials determine which public corpus and private user scopes are available. Read, write, sharing, and destructive actions remain distinct.

  3. 03

    Retrieval finds stored evidence

    Lexical and semantic search, metadata filters, curated relationships, and source-aware reranking locate exact passages. Missing coverage stays visible rather than being filled with generated text.

  4. 04

    The agent receives structured context

    Responses return evidence in labeled fields with stable IDs, provenance, caveats, confidence, and citation data so the client can display, verify, or continue working with it.

Typed, scoped tools

What developers can build with it.

Each capability is exposed as a typed MCP tool with a bounded job. Clients can combine them into research, study, writing, and verification workflows.

Find

Search exact sources

Inspect the active corpus release and verification coverage, search stored passages in lexical, semantic, or hybrid mode, then fetch a source or expand a passage into its immediate stored context.
get_corpus_manifestsearch_sourcesget_sourceget_passage_context

Understand

Build Scripture context

Retrieve an exact passage alongside literary context, corpus-derived cross-references, concepts, patristic commentary, and conciliar material—kept in separate fields.
get_scripture_contextget_primary_sources

Trace

Follow concepts and words

Trace curated links among Scripture, theologians, councils, controversies, doctrines, and passages. Look up biblical lemmas, morphology, corpus-counted renderings, occurrence samples, and translation cautions.
trace_conceptlookup_theological_termlookup_biblical_word

Compare

Preserve real differences

Compare explicitly attributed tradition positions and separate consensus, disagreement, and missing coverage. Evaluate a proposition against curated supporting, contradicting, qualifying, and contextual evidence.
compare_traditionsevaluate_claim

Verify

Resolve and check citations

Resolve references, check one quotation or a batch of up to 20 against immutable release snapshots, and format citations. Results distinguish exact matches, normalized matches, missing evidence, and operational errors.
resolve_citationverify_citationverify_citationsformat_citation

Assemble

Build inspectable evidence

Assemble sources, Scripture, attributed positions, provenance, and caveats into an evidence packet. For curated claims, inspect an evidence graph with disclosed rules and an unsigned integrity commitment; it does not certify theological truth.
build_evidence_packetbuild_claim_proof

Remember

Continue a private study

With user-approved scopes, search personal study context and continue from saved passages, notes, progress, plans, and authored reflections without allowing that data to rewrite canonical results.
search_personal_contextcontinue_studyget_study_progress

Research

Run a durable project

Save evidence immutably, pin or exclude passages, annotate findings, export the project, fork a line of inquiry, and manage revocable read-only shares.
save_research_evidencecurate_research_evidenceexport_research_project

Example workflow

Ask for evidence,
not a verdict.

A client can ask Feason to assemble the relevant material, then let its own model analyze the structured response. Feason does not publish a final theological judgment on the model’s behalf.

Explore six workflows →

Tool callbuild_evidence_packet
{
  "question": "Trinity",
  "traditions": ["Catholic", "Eastern Orthodox", "Reformed"],
  "limit": 12
}
Response fields (excerpt)
{
  "primarySources": [...],
  "scripture": [...],
  "historicalContext": [...],
  "interpretivePositions": [...],
  "sourceProvenance": [...],
  "retrievalMetadata": {
    "corpusVersion": "...",
    "caveats": [...]
  }
}

Use cases

Expand what your agent can help with.

These workflows combine existing tools. Start by inspecting the corpus manifest and the tools your connection exposes; private workflows also require a user-bound grant.

Review an article’s citations

Give editors a quotation-by-quotation report with source references and a bibliography.
Prompt and workflow

Try asking: “Check the theological quotations in this draft against the active Feason corpus release. Separate exact matches, normalized matches, unresolved citations, and errors; format resolved citations in Chicago style.”

Tool sequence

  1. get_corpus_manifest
  2. resolve_citation
  3. verify_citations
  4. format_citation

Minimum scopes: context:read

Resolve passage IDs first, then verify batches of at most 20 with the release version. A normalized match is not an exact quotation; unresolved text may be outside corpus coverage.

Prepare a teaching brief

Build a source-backed starting point for a sermon, lesson, or discussion of Scripture.
Prompt and workflow

Try asking: “Prepare an evidence brief on Ephesians 2:8. Keep Scripture, historical sources, and attributed interpretations separate; preserve retrieval warnings and check quotations from stored corpus passages.”

Tool sequence

  1. get_scripture_context
  2. build_evidence_packet
  3. verify_citations

Minimum scopes: context:read

Check that the returned reference matches the requested passage and retain any fallback warning. The client drafts the lesson; disputed readings stay attributed to their sources.

Explore a biblical word

Support language study with attributed lexical data and examples of usage in context.
Prompt and workflow

Try asking: “Look up the original-language words in John 1:1, then examine logos using the returned lemma or Strong’s code. Show morphology, occurrence samples, and translation cautions.”

Tool sequence

  1. get_scripture_context
  2. lookup_biblical_word

Minimum scopes: context:read

Each word lookup takes one verse, Strong’s code, or term. Samples are bounded, coverage may be missing, and a word’s root or gloss alone does not settle its meaning in a verse.

Prepare a tradition seminar

Compare named witnesses while keeping disagreement and missing coverage visible.
Prompt and workflow

Try asking: “Compare Baptist, Lutheran, Anglican, and Reformed sources on baptism. Attribute each position to its exact witness, show the surrounding passages, and list questions the corpus cannot answer.”

Tool sequence

  1. compare_traditions
  2. get_passage_context
  3. format_citation

Minimum scopes: context:read

A witness does not represent every member of a tradition. Missing coverage is not evidence that a tradition has no teaching, and agreement is limited to the sourced assertions returned.

Keep a research bibliography

Preserve evidence for a private inquiry and export it for writing or reference management.
Prompt and workflow

Try asking: “Create a private project on early Christian accounts of resurrection, save an evidence packet, and export the project as Markdown and BibTeX with its source references.”

Tool sequence

  1. create_research_project
  2. save_research_evidence
  3. export_research_project

Minimum scopes: context:read, study:write

Use the returned project ID. Saving retrieves a fresh packet; it does not upload arbitrary evidence. Exports preserve saved evidence, while sharing is a separate action.

Resume a member’s learning

Combine authorized study history and lecture notes into a focused review session.
Prompt and workflow

Try asking: “Resume my study of grace using my saved study context and lecture notes. Suggest three review questions and retrieve canonical evidence separately from my personal notes.”

Tool sequence

  1. continue_study
  2. get_lecture_notes
  3. build_evidence_packet

Minimum scopes: context:read, study:read, lecture:read

Private tools require a bound Feason account and explicit scopes. Journal text is unavailable. The client proposes review questions; these calls do not update course or memorization progress.

Designed for responsible use

The trust model is part of the product.

Feason is designed to make an agent’s evidence easier to inspect while limiting what the integration can access and what the system retains.

Evidence stays attributable

Stored passages carry their source provenance. Curated relationships and generated summaries are not presented as source text.

Traditions stay named

Comparisons attribute positions and expose disagreement or missing corpus coverage instead of collapsing them into a single answer.

Access stays scoped

OAuth consent and revocable client credentials gate private sources. An integration receives only the scopes a user grants.

Private stays separate

Journal titles and bodies are client-side encrypted and cannot be decrypted or searched by MCP. Other private context never changes canonical ranking.

Operations stay auditable

Feason records tool outcome, client, duration, and request identifiers for reliability and abuse prevention—not full requests, access tokens, or query text.

Quality stays measurable

Corpus coverage, citation accuracy, source drift, retrieval health, conformance, latency, failures, and human review are monitored as separate signals.

Connect

Use Feason from the AI client you already work in.

MCP requires a credential with context:read before its first request. OAuth-capable clients prompt you to sign in when connecting; private tools require additional explicit scopes.

Connect

Codex

Add the remote HTTP server to your Codex configuration, then sign in through Feason OAuth to grant context:read before connecting.
[mcp_servers.feason]
url = "https://www.feason.com/api/mcp"

Connect

Claude

Add a custom remote MCP connector, enter the endpoint, and approve only the Feason scopes your workflow needs.
https://www.feason.com/api/mcp

Connect

ChatGPT

Create a custom MCP connector with the endpoint, then complete the Feason authorization and consent flow.
https://www.feason.com/api/mcp

Public evidence tools require context:read and use only the shared canonical corpus.

Private tools require a signed-in Feason account and the relevant read or write scope.

Manage connected apps →

Essential storage keeps you signed in and remembers preferences. With your permission, Feason also uses first-party analytics and Google Ads conversion measurement to learn whether a campaign led to a useful action. Google measurement is disabled until you accept, and we disable ad personalization and remarketing. See our Privacy Policy for details.