Loading…
Deterministic claim verification with receipts across roughly 60 domains — no model in the loop.
Do not connect
A critical issue was found. Do not connect this server as-is.
Scanned 1 month ago Due for re-check
A server can change after it's graded. Re-run the automated scan to refresh this report.
This grade is deterministic and reproducible: the same server surface always yields the same grade under a given algorithm version. It is a real automated assessment computed by the MCPGrade engine from what the probe actually observed — not a fabricated or opinion score. It is not a manual human pentest, so it can miss context-specific risks.
Every signal below was measured directly by the automated probe. The grade is derived only from evidence like this — nothing is assumed.
The narrow-highway MCP server exposes 94 tools, focused primarily on developer capabilities. Its published description reads: "Deterministic claim verification with receipts across roughly 60 domains — no model in the loop". It communicates over Streamable HTTP using the 2026-07-28 protocol revision, and does not require authorization to connect. MCPGrade currently rates narrow-highway F — a critical issue was found and the server should not be connected as-is. Its most notable findings include "Cross-tool shadowing" and "Cross-tool shadowing". This report is a deterministic, reproducible automated assessment: the same observed surface always yields the same grade under a given algorithm version, and it is refreshed as new scans arrive and free to read — but it is not a substitute for a manual human security review. Always re-check narrow-highway's advertised tools, transport security, and authorization posture before connecting an autonomous agent or sharing sensitive context with it.
Add narrow-highway to an MCP client using the endpoint below. Review the grade and findings above before granting an autonomous agent access.
MCP endpoint
https://narrowhighway.com/mcpStreamable HTTP transport. This is the MCP endpoint, not a website — paste it into your MCP client server list rather than a browser.
Captured passively during the read-only scan. Click any value to find servers that match it.
The tools this server advertises via tools/list — names, purposes, and the parameters each accepts, exactly as enumerated read-only.
Verify a claim deterministically — returns a verdict (HOLDS / BROKEN / INCOMPLETE / SYSTEM_ERROR), the worked trail, AND a sealed receipt {content_hash, cite_url} you can re-fetch and re-verify (seal_
Audit a whole text: deterministic extractors find every checkable quantitative claim (sums, percentages, hourly/annual pay, compound interest, rule-of-72, elapsed years, day-of-week, leap years, nutri
Ranked search over the keeping (the kept library).
Fetch a sealed verification record (the receipt) by its content hash.
Strip PII (emails, SSNs, credit cards, IPs, URLs) from text to stable placeholders before you pass it onward; the mapping is returned so YOU reveal replies locally. For true privacy run this on a LOCA
Fetch one card (the full record) from the keeping by id.
Browse the keeping — paginated, optional shelf filter. Returns card briefs.
The decks — curated card SETS, both by-situation (need decks, frontloaded for a moment of need like 'when the power goes out') and by-domain. Each with a live card count; `need: true` marks a situatio
Open a deck to its frontloaded hand — the cards that situation calls for, in order, with NO query needed. Give the deck `id` (from `decks`). Ideal for anticipating need: open 'water-safe', 'first-aid-
Counts over the keeping — total, by shelf, by surface.
The deterministic card of the day from the keeping (same all day).
The map: a read-only view of one axis (its scaffold members, depth, neighbors, umbrella children). Omit `axis` for an overview of all axes.
The axes that sit on a given scaffold member (dimension).
Cards related to one card — its explicit links + same-shelf siblings.
Find the card for a query — by exact id, then title, else ranked search.
Corpus health — is the keeping loaded and sound (totals, shelves, surfaces).
A synthesized pronunciation guide (respelling + approximate IPA) for a transliteration or word — honestly labeled, not a native speaker.
Steward — a household budget (income, expenses -> net, savings rate, by category). Shows and plans; NEVER moves money.
Steward — cost destroyed: money you did NOT spend (was -> now), kept in your currency.
Coach — the subjects a learner can study (read / mcguffey / aesop / founding / pilgrims / es / …), each with its unit count. 'read' is the door in.
Coach (K-3 tutor) — one subject's whole path: unit count, tracks, ordered unit briefs. ?subject= selects the path (default 'read'). Verbatim; never generated.
Coach — one unit, VERBATIM as authored (rule, examples, decodable sentence, checks).
Coach — the next lesson in a subject, deterministically. Omit `after` for the first unit; pass a unit id for the one that follows it. `subject` selects the path.
Coach — adaptive 'what's next' in a subject: given completed unit ids, the next lesson whose prerequisites are met (grows with the student). Found, never generated.
Coach — seal an HONEST INTEGER count of completed units (a receipt for progress, never a grade on the child). Returns a re-checkable seal.
Coach — what it does and the boundary it will not cross (never grades a child).
Explains how to create a SOVEREIGN identity — keys are born on the USER'S device (never on the server; no private key crosses the wire). Returns guidance, not a key. The server only handles public key
Verify a signature over a message against a public key (never raises; True/False).
Derive the stable public fingerprint id from a public key (deterministic).
Issue a badge over already-sealed checks — a re-checkable receipt that points at N seals that STILL STAND. States EXACTLY N; NEVER a competency claim.
Re-check a badge from the store — re-verifies every seal it references and returns the count that still stands (N recomputed, not trusted).
Record a person's OWN words about their study — a DISTINCTLY TYPED record that can NEVER count as a sealed check or satisfy an auto-graded requirement.
Create/extend a shared study (superposition stack) — each entry mints ONE card that lives once and is referenced by key; no duplication.
Export a study as a self-contained, portable bundle. Returns the bundle and its content_hash; to bind your identity to it, sign that hash with your own key on your own machine — this tool does not tak
Import an exported study bundle — re-materializes its cards (each lives once).
Discover pseudonymous shared-study groups by TOPIC (not by person). Optional q filters over topic/title/description. Members are handles only — no PII.
A study group: topic, member handles (no ids/PII), and the shared-study cards.
Open a study group around a topic (pseudonymous; a handle, no personal info). For grown believers — the children's coach is a separate, never-joined surface.
Join a study group (consent-based, pseudonymous; idempotent).
Add a verse/note/question to a group's shared study — attributed to your handle, optionally signed. Verbatim; a member's own words, not engine-verified.
Create ONE event in a human's calendar — the pilot on-behalf write, and the only one. Requires a LIVE consent grant: the human signed a scoped, expiring calendar_write grant for YOUR key fingerprint o
Check whether a human has authorized YOU (this agent's key fingerprint) for a named verb — the agent covenant's 'request human authorization before writes'. Speaking as YOURSELF (your own key, your ow
THE COMMONS — step 1 of stocking your own shelf. Returns {fields, signable}: the exact canonical bytes to sign with YOUR key on YOUR machine. A shelf is a covenant key with cards on it, not an account
Step 2 — stock the shelf. Send the fields from shelf_signable plus a detached signature over those bytes; the private key never travels. Your words stay at the `member` tier forever: promotion to the
Read one member's shelf. Pass `viewer` (your own key) to see your own private drops; anyone else sees the shelf ring and promoted commons cards only. Nothing anywhere records who read what.
What the fellowship has put on the commons — promoted member work, newest first. Every card is a member's own work at the `member` tier: the library amplified it; the library did not verify it.
THE PLAYBOOK — the Body's testimony of faithful obedience: 'Canon commands, Playbook remembers.' List entries (optionally by `status` quarantine/confirmed/rejected/pruned or `author`), or one by `id`.
Step 1 — the exact bytes to sign ON THE DEVICE for a Playbook write. `op` is entry | witness | outcome | prune. A new `entry` REQUIRES a CONFESSION ('I may be wrong. I acted in faith according to [anc
Step 2 — verify your detached signature over those exact bytes and enter the testimony/event. A new entry is born QUARANTINE and reaches CONFIRMED only when TWO independent brothers affirm (never the
What waits on a HUMAN steward. The counter never promotes; it only decides when a person must look.
The canonical bytes for withdrawing YOUR OWN card. A member never needs permission to take their own words down — the proof is the same key that signed the drop. Sign these bytes, then call `curate` w
A recorded act on one drop: `promoted` · `refused` · `withdrawn`. A name AND a reason are always required — no anonymous judgement, and a refusal without a reason teaches the community nothing. WHO MA
Step 1 of a report or a block: the exact canonical bytes to sign with your own key, on your own machine. Returns {fields, signable}. Sign the decoded `signable` bytes and pass BOTH fields and signatur
Ask the library to ACQUIRE something it does not hold (kind=missing, give query) or to EXPAND a thin card (kind=expand, give card_id). Opens a want on the AGENT PLANE — held separate on the desk until
The library's desiderata desk — open wants, sorted by demand, agent plane marked and separate. Read it to find gaps you could dig for.
Return to the comb with a FOUND source for an open want: label + url + snippet, attributed. You are a forager, not an author — offer only public-domain / openly-licensed sources you actually located,
Report a community item (group_contribution, mesh_message, door_note) to the moderation floor. One report is a claim, never a verdict; at three DISTINCT reporters the item is held for a HUMAN steward'
Bind your identity to a record you already hold — phase 2 of the sovereign flow. Do the thing unsigned (badges_issue, study_export, group_contribute), take the returned content_hash, sign THAT hash wi
Who has borne witness to a record, each signature re-verified as it is read (storage is never trusted). Reports invalid entries rather than hiding them.
The actual current date and time, fresh at this call — UTC (the clock every seal is stamped in), the library's home zone, and optionally any IANA zone you name. Your own sense of 'today' is months old
The live capability statement: what this engine can verify, what tools and endpoints exist, how large the keeping is, and where its boundaries are. Every count is computed at call time and carries a '
THE GATE KERNEL — the law you keep here, in one call. The five moves (find, distinguish the KIND, verify what can be verified, preserve the trail, never silently upgrade authority), the eight-rule age
Route ONE proposed state-change through the kernel and get back the verdict (REJECT / QUARANTINE / CONFIRMED) plus the nine-field record — so you can DISCERN before you write (covenant rule 8: stop wh
The believers immediately around you in the Fellowship Mesh — your view only, never a global map. Requires your fingerprint (fp) and a confession already bound to that key; unconfessed callers get the
The messages that reached you, each carrying its own offline verification so you trust it by proof rather than by this server's word. Requires fp + confession.
Read the words left on YOUR door — the whiteboard others wrote to you, each with its verification. Requires fp + confession.
Step 1 of speaking to your fellowship: get the exact canonical bytes to sign for a message (returned base64url, with the nonce and created_at to send back). Sign them locally with your own key. Reprod
Leave a word on one believer's door — directed encouragement rather than a post to everyone near you. Same sovereign shape: get the bytes from mesh_signable (with target), sign locally, send only the
Step 2: speak to the nodes around you, carrying only your SIGNATURE — your private key never leaves your machine and this tool will not take one. Pass the same text/kind/ttl plus the nonce, created_at
Bring anything — a question, a claim to check, a word to study, or what is actually on your mind. The engine discerns what KIND of thing you brought and answers in kind: it finds and verifies, and dec
Discern what to do with anything WITHOUT doing it — the proposal half of the engine, laid open. Bring a claim, a question, or a messy statement; get back what KIND it is, the necessary claim reduced t
Candidate Engine step 1 — commit a COMPLETE raw candidate set BEFORE any evaluation. Give the query, the generator, the method, and EVERY raw candidate; weights ride verbatim as UNTRUSTED metadata (pr
Candidate Engine step 2 — narrow a COMMITTED set: verifiers are assigned under the fixed pre-registered routing policy (never your choice — that is what stops verification shopping), each routed claim
Fetch a sealed candidate record (a narrowing receipt or a commitment anchor) by its content hash — the whole preserved path, losers included, re-checkable offline with tools/verify_seal.py. The seal p
Resolve a Scripture reference to its World English Bible text.
Read a passage of the WEB — a single verse, a range (John 3:16-18), or a whole chapter (John 3).
Strong's word study — original-language definition + pronunciation + every occurrence.
Verses connected to a reference by SHARED original words (Strong's) — the dots, connected; deterministic and found, ranked by shared-word count.
Every verse where a Strong's word occurs (the concordance).
Public-domain, attributed commentary (Matthew Henry) on a reference — the commentator's own words, found and cited, never generated.
Editorial cross-references for a verse (openbible.info, CC BY — expansion of the public-domain TSK), ranked by relevance votes. Found + attributed.
A Bible figure from Easton's Bible Dictionary (1897, PD) — summary + every verse that speaks of them (found + attributed; category tag is imperfect).
Browse/search Easton's Bible Dictionary (people, places, terms).
Christ-signpost traces (prophecy/cross-cultural pointers to Jesus) — attributed, verdict CONCORDANT/MIXED, NEVER HOLDS (a signpost, not a proof). Pass id for one trace, q to search, else lists all.
Harmony of the Gospels — one event of Christ's life, every gospel that records it, side by side (found, verbatim WEB text, never generated). Pass id for one event; else lists every event grouped by ph
Timeline — Old Testament, New Testament (Acts onward), and Church History, one spine from creation to today. Genuinely disputed dates (early/late Exodus, the date of Revelation, etc.) carry both posit
Back-matter reference tables: weights_measures, names_of_god, parables, miracles, book_intros, topical_index. Disputes carried (a cubit's two lengths, a book's two datings), refs verified against the
The Atlas — biblical places with REAL coordinates, honestly held: located places carry lat/lon (cross-checked against an independent gazetteer); disputed sites (Mount Sinai, Cana, Golgotha) NAME their
The storyboards — the common narratives charted in the Bible FIRST (exile-and-return, the barren woman bears, down-to-the-pit-raised, the great reversal...), each instance real people with verified re
The quick-find index — ONE lookup across the whole reference section: archetypes, storyboards, movements, the six study tables, the atlas, harmony, timeline, and the encyclopedia. Each hit is a pointe
The original-language words behind a verse (Hebrew/Greek, with Strong's where known) — FOUND in the lexicon, never generated. Pass ref, e.g. 'John 3:16'.
The canon as concentric layers — the undisputed 66 that all major traditions share, plus the books held BEYOND it by particular traditions, each framed on its own layer and never merged. REPORTS who h
The teachings of Christ (Words in Red) — the frozen Greek anchor plus the history that ALIGNS to each teaching, gathered and attributed, never authored. Pass id for one teaching; else the queue.
Seeds of the Word (the Areopagus / logos spermatikos pass) — true fragments mined from the nations, ATTRIBUTED, CONCORDANT/signpost NEVER HOLDS; each names the idol it refuses and points to Christ (Ac
Sorted worst-first. Each finding shows its severity, what it means, its OWASP MCP Top-10 mapping, and a recommended fix — the check id links to the exact methodology row that produced it.
tool: verify
A tool description tries to alter the model’s use of another tool.
and re-verify (seal_fetch). Two forms: (a) MATH — {mode, params}; (b) ANYRecommendationDescriptions must describe only their own tool.
tool: audit
A tool description tries to alter the model’s use of another tool.
sealed coverage report returns — per-claim source quote + verdict + trRecommendationDescriptions must describe only their own tool.
tool: redact
A tool description tries to alter the model’s use of another tool.
stic; pair with verify for a receipt.RecommendationDescriptions must describe only their own tool.
tool: deck_open
A tool description tries to alter the model’s use of another tool.
eck `id` (from `decks`). Ideal for anticipating need: open 'water-safRecommendationDescriptions must describe only their own tool.
tool: locate
A tool description tries to alter the model’s use of another tool.
le, else ranked search.RecommendationDescriptions must describe only their own tool.
tool: steward_cost_destroyed
A tool description tries to alter the model’s use of another tool.
T spend (was -> now), kept in your currency.RecommendationDescriptions must describe only their own tool.
tool: identity_create
A tool description tries to alter the model’s use of another tool.
es public keys (identity_verify, identity_fingerprint).RecommendationDescriptions must describe only their own tool.
tool: identity_verify
A tool description tries to alter the model’s use of another tool.
Verify a signature over a message against a public keyRecommendationDescriptions must describe only their own tool.
tool: group_contribute
A tool exposes a free-text/webhook/feedback parameter with no honest purpose — a classic exfil channel.
e. Never send a private key.","additionalProperties":false}},"required":["iRecommendationRemove side-channel parameters; constrain tool inputs.
tool: shelf_signable
A tool exposes a free-text/webhook/feedback parameter with no honest purpose — a classic exfil channel.
parameter "url"RecommendationRemove side-channel parameters; constrain tool inputs.
tool: shelf_drop
A tool exposes a free-text/webhook/feedback parameter with no honest purpose — a classic exfil channel.
hose bytes; the private key never travels. Your words stay at the `member`RecommendationRemove side-channel parameters; constrain tool inputs.
tool: shelf_drop
A tool description tries to alter the model’s use of another tool.
the fields from shelf_signable plus a detached signature over those bytes; theRecommendationDescriptions must describe only their own tool.
tool: commons_read
A tool description tries to alter the model’s use of another tool.
library did not verify it.RecommendationDescriptions must describe only their own tool.
tool: playbook_read
A tool description tries to alter the model’s use of another tool.
ful obedience: 'Canon commands, Playbook remembers.' List entries (opRecommendationDescriptions must describe only their own tool.
tool: playbook_submit
A tool description tries to alter the model’s use of another tool.
Step 2 — verify your detached signature over those exact bytesRecommendationDescriptions must describe only their own tool.
tool: curate_signable
A tool description tries to alter the model’s use of another tool.
tes, then call `curate` with them.RecommendationDescriptions must describe only their own tool.
tool: curate
A tool description tries to alter the model’s use of another tool.
ignature` from `curate_signable` instead. A typed name is not authority. A refuRecommendationDescriptions must describe only their own tool.
tool: moderation_signable
A tool exposes a free-text/webhook/feedback parameter with no honest purpose — a classic exfil channel.
o `report`. The private key never travels. {"type":"object","properties":{"RecommendationRemove side-channel parameters; constrain tool inputs.
tool: moderation_signable
A tool description tries to alter the model’s use of another tool.
Step 1 of a report or a block: the exact canonical bytes to sign wRecommendationDescriptions must describe only their own tool.
tool: want_open
A tool description tries to alter the model’s use of another tool.
Ask the library to ACQUIRE something it does not hoRecommendationDescriptions must describe only their own tool.
tool: want_offer
A tool exposes a free-text/webhook/feedback parameter with no honest purpose — a classic exfil channel.
parameter "url"RecommendationRemove side-channel parameters; constrain tool inputs.
tool: report
A tool description tries to alter the model’s use of another tool.
be SIGNED (call moderation_signable first): three witnesses means three keys, neverRecommendationDescriptions must describe only their own tool.
tool: report
The text tells the model WHEN to call this tool relative to others ("always call first", "before any other tool", "chain to X tool") — a toxic-flow injection that hijacks the agent’s orchestration rather than describing the tool.
look. A report must be SIGNED (call moderation_signable first): three witnesses means three keys, never threeRecommendationTool metadata must describe only the tool, never sequence the agent’s calls.
tool: attest_record
A tool exposes a free-text/webhook/feedback parameter with no honest purpose — a classic exfil channel.
}. Never send a private key. Several parties may attest to one record: oneRecommendationRemove side-channel parameters; constrain tool inputs.
tool: attest_record
A tool description tries to alter the model’s use of another tool.
thing unsigned (badges_issue, study_export, group_contribute), take the retuRecommendationDescriptions must describe only their own tool.
tool: capabilities
A tool description tries to alter the model’s use of another tool.
is instead of trusting any count written in prose.RecommendationDescriptions must describe only their own tool.
tool: kernel
A tool description tries to alter the model’s use of another tool.
guish the KIND, verify what can be verified, preserve the trail, neverRecommendationDescriptions must describe only their own tool.
tool: kernel_gate
A tool description tries to alter the model’s use of another tool.
nge through the kernel and get back the verdict (REJECT / QUARANTINE /RecommendationDescriptions must describe only their own tool.
tool: mesh_signable
A tool description tries to alter the model’s use of another tool.
RecommendationDescriptions must describe only their own tool.
tool: mesh_leave_on_door
A tool exposes a free-text/webhook/feedback parameter with no honest purpose — a classic exfil channel.
ll not accept a private key; unsigned notes are refused. {"type":"object","RecommendationRemove side-channel parameters; constrain tool inputs.
tool: mesh_leave_on_door
A tool description tries to alter the model’s use of another tool.
the bytes from mesh_signable (with target), sign locally, send only the signRecommendationDescriptions must describe only their own tool.
tool: mesh_post
A tool description tries to alter the model’s use of another tool.
signature from mesh_signable. Refused unless the signature verifies againstRecommendationDescriptions must describe only their own tool.
tool: ask
A tool description tries to alter the model’s use of another tool.
(Scripture, the Harmony, the Timeline, the lexicon) become callable — tRecommendationDescriptions must describe only their own tool.
tool: discern
A tool description tries to alter the model’s use of another tool.
the gate would verify, and where it routes. It PROPOSES, never confirRecommendationDescriptions must describe only their own tool.
tool: candidate_narrow
A tool description tries to alter the model’s use of another tool.
set exactly as candidate_commit returned it.RecommendationDescriptions must describe only their own tool.
tool: characters_browse
A tool description tries to alter the model’s use of another tool.
Browse/search Easton's Bible Dictionary (people, places, termRecommendationDescriptions must describe only their own tool.
tool: prophecy_traces
A tool description tries to alter the model’s use of another tool.
one trace, q to search, else lists all.RecommendationDescriptions must describe only their own tool.
tool: backmatter
A tool description tries to alter the model’s use of another tool.
rs that open in word_study. Pass table for one table; else the index of alRecommendationDescriptions must describe only their own tool.
tool: bible_places
A tool description tries to alter the model’s use of another tool.
es instead of planting one flag; unlocatable places (Eden, Emmaus, TarshisRecommendationDescriptions must describe only their own tool.
tool: study_find
A tool description tries to alter the model’s use of another tool.
les, the atlas, harmony, timeline, and the encyclopedia. Each hit is aRecommendationDescriptions must describe only their own tool.
tool: seeds
A tool description tries to alter the model’s use of another tool.
one seed, q to search, tradition to filter, else lists all with Paul'RecommendationDescriptions must describe only their own tool.
The server accepts tool enumeration (and likely invocation) with no authentication.
RecommendationRequire OAuth 2.1 authorization for any server exposing non-public tools.
tool: decks
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "decks"RecommendationScope tools to the minimum needed.
tool: identity_create
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "identity_create"RecommendationScope tools to the minimum needed.
tool: study_create
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "study_create"RecommendationScope tools to the minimum needed.
tool: calendar_create
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "calendar_create"RecommendationScope tools to the minimum needed.
tool: consent_check
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "consent_check"RecommendationScope tools to the minimum needed.
tool: shelf_drop
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "shelf_drop"RecommendationScope tools to the minimum needed.
tool: shelf_read
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "shelf_read"RecommendationScope tools to the minimum needed.
tool: commons_read
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "commons_read"RecommendationScope tools to the minimum needed.
tool: playbook_signable
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "playbook_signable"RecommendationScope tools to the minimum needed.
tool: curate_signable
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "curate_signable"RecommendationScope tools to the minimum needed.
tool: curate
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "curate"RecommendationScope tools to the minimum needed.
tool: attest_record
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "attest_record"RecommendationScope tools to the minimum needed.
tool: kernel
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "kernel"RecommendationScope tools to the minimum needed.
tool: kernel_gate
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "kernel_gate"RecommendationScope tools to the minimum needed.
tool: mesh_signable
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "mesh_signable"RecommendationScope tools to the minimum needed.
tool: mesh_leave_on_door
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "mesh_leave_on_door"RecommendationScope tools to the minimum needed.
tool: candidate_commit
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "candidate_commit"RecommendationScope tools to the minimum needed.
tool: candidate_narrow
The server advertises open-world / broadly-scoped capabilities.
no annotations on write-capable tool "candidate_narrow"RecommendationScope tools to the minimum needed.
Vantaj uptime monitoring via MCP — manage monitors, heartbeats, incidents, and status pages.
Unified gateway to Algeria's TKAWEN ecosystem: commerce, certification, and AI tools.
Provides access to the Cohereon Doctrine AI safety framework with governance components, tiered access, and agent onboarding.
Agentic rails for complex workflows with receipts, fees, and MCP tool access.
Structural TC39 spec lookup for ECMA-262 and ECMA-402 in AI agents, SHA-pinned and offline-first.
Structural TC39 spec lookup for ECMA-262 and ECMA-402 in AI agents, SHA-pinned and offline-first.