MCP server

Install

One click to add Entidex to your editor or AI client. Hosted at https://entidex.com/api/mcp — nothing to install or self-host. Replace entx_live_sk_... with your API key after connecting.

Or from the terminal — npx @entidex/mcp-setup detects your client and writes the config:

npx @entidex/mcp-setup                        # interactive — detects your client, prompts for a key
npx @entidex/mcp-setup --client claude-code   # non-interactive: claude-desktop | claude-code | cursor | vscode
npx @entidex/mcp-setup --demo                 # public demo key (reference cohort — no account needed)
npx @entidex/mcp-setup --uninstall            # remove Entidex from a client config

Manual config

If your client doesn’t support one-click install, paste the config directly. All clients need the endpoint URL and a Bearer token; stdio-only hosts (Claude Desktop) bridge it with mcp-remote.

{
  "mcpServers": {
    "entidex": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote",
        "https://entidex.com/api/mcp",
        "--header", "Authorization: Bearer ${ENTIDEX_API_KEY}"
      ],
      "env": { "ENTIDEX_API_KEY": "entx_live_sk_..." }
    }
  }
}

Tools exposed

28 tools. Each enforces the same access level as its REST equivalent — tools you can’t call still appear in tools/list so LLMs can discover the ladder; a paid tool called on a free key returns a clear tier error.

Free — no signup

• entity_complete — the complete, categorized intelligence record in one call: identity, categorization (type, lifecycle stage, surfaces, topics, cohorts), headline signals (AI Visibility, Sentiment, Share of Voice, Trust Stack), surface coverage, scores, perception, derived insights, drift, lifecycle, relationships, decision survival, and cohort. Free callers get a generous taste; higher tiers unlock depth additively, with withheld sections listed in meta.withheld.
• entity_resolve — resolve a raw string to a canonical Entidex ID with a confidence band.
• creator_resolve — resolve a creator / podcast name or any one profile link to the single canonical creator + every cross-verified platform handle with provenance + confidence + a link to the signed record. Three-state verdict: confirmed / ambiguous / not_found.
• entity_grounding_packet — the Grounding Packet: one small, deterministic per-entity grounding payload over the verified record — facts with provenance, corrections to common AI claims, and Ed25519-signed provenance. Fetch it before answering about an entity.
• entity_grounding_check — check a statement against the verified record BEFORE asserting it. Returns a coded verdict per extracted assertion (supported / contradicted / unverifiable) with confidence and a citation into the signed Knowledge Statement.

Starter+

• entity_verify_fact — verify facts against the Entidex verified record: verified values with source and date, plus where AI engines currently diverge (the correction delta) and signed provenance.
• entity_handles — every social handle for an entity, each with platform, canonical URL, confidence and provenance tier. The reverse-lookup half of cross-platform identity resolution.
• entity_markets— latest stock-market reading (ticker, exchange, price, % change, provenance) — the entity’s own listing or that of its public owner / parent. Null when not publicly traded.
• entity_ecosystem — cross-engine AI Ecosystem read: per-engine fit, citation rates, recommendation patterns, and the strategic narrative grounded in observed platform signals.
• entity_score_history — time-series snapshots over AI Visibility, Sentiment, Share of Voice, Trust Stack, Narrative Coverage, or Entity Maturity, with current / previous / drift. Powers sparklines and drift narratives.
• entity_consensus— cross-surface consensus / divergence snapshot: how much surfaces agree on the entity’s standing, with optional over-time trend matrix.
• entity_lifecycle — lifecycle stage classifier (emergence through resurrection) with drivers and predicted next stage.
• entity_knowledge_accuracy — how accurately each AI engine actually knows the entity vs the verified record. Overall accuracy % with 95% Wilson interval + per-engine breakdown of the exact fields each engine gets wrong. Zero LLM cost.
• entity_recommendation_potential — per-prompt-class breakdown of how likely AI engines are to recommend the entity (discovery / evaluation under pushback / constrained choice), with per-engine rate + 95% Wilson interval.
• entity_guest_intelligence — per-episode cast read for a show / creator: video sentiment, per-role sentiment with sample size, top praise / criticism, Claim Truth verdicts, plus a Guest Impact roll-up (which appearances moved AI Visibility 7d pre → 7d post).
• entity_episode_adoption_lag — per-episode Episode → Engine Adoption Lag: for each recent episode, how long until each of the five canonical AI engines can name it.
• entity_serp_vs_ai — SERP vs AI presence comparison with a coded gap verdict (aligned / serp_dominant / ai_dominant / both_strong / both_weak).
• entity_milo_action — role × lens action-shaped read grounded in every dataset surface. Pick who the read is for (sales / executive / communications / investor / …) and what they want (situation-brief / meeting-prep / competitor-view / truth-gap / crisis-watch / …).
• entity_submit_feedback — submit a correction or feedback back to Entidex. Captured append-only with provenance and reviewed before it affects the published record.

Pro+

• entity_intelligence — the layered intelligence view (profile + scores) for a resolved entity.
• entity_core_intelligence — Entity Core Intelligence Profile: ten authority / readability scores (clarity, authority, trust, entity graph, topical authority, recommendation, market clarity, machine readability, momentum, reputation safety) plus topic ownership, entity graph, competitors, perception map, strategic actions, and live-web grounding provenance.
• entity_drift — drift events over the tracked window (named events and cross-surface movement).
• entity_timeline — unified chronological timeline: one dated, most-recent-first stream interleaving score movements, drift events, lifecycle transitions, X inflection + amplification, Claim Truth transitions + narrative divergence, and captured AI-engine responses + Decision Survival runs.
• entity_decision_survival — latest Decision Survival result: per-scenario survival rate and the displacement chain (which competitors took the recommendation slot).
• entity_knowledge_analysis — what AI engines get wrong, miss, or disagree on. Compact summary by default; use section or detail for per-field data.
• entity_x_sentiment — latest X (social) sentiment: weighted polarity, key voices, praises / complaints, narrative context, early-warning spikes, and citation-backed source posts.

Agency

• entity_cohort — cohort position: which cohorts the entity belongs to, rank, and deltas vs cohort median on AI Visibility, Sentiment, Share of Voice.
• entity_cohort_rankings — ranked leaderboard for a cohort: members ranked by Share of Voice (or visibility / sentiment).

Canonical correction channel

Entidex is the canonical place to correct what AI gets wrong about an entity. Read the verified record with entity_verify_fact; if it’s wrong, submit a correction with entity_submit_feedback. Corrections are captured append-only with provenance and reviewed before they affect the published record — so the channel stays trustworthy.

Want to evaluate without an account? Run npx @entidex/mcp-setup --demo — it configures the public demo key, which grants read + correction access to a small reference cohort with nothing to provision. Machine-readable descriptor at /.well-known/mcp.json.

Open source & listings

The MCP server config, examples, and setup tooling are all open source. Use them as-is or fork to build your own integration.

• @entidex/mcp-setup — one-command installer on npm. npx @entidex/mcp-setup detects your client and writes the config.
• com.entidex/entidex — listed on the official MCP Registry. Streamable-HTTP server descriptor with bearer-token auth and DNS-verified namespace.
• Integration recipes — ready-to-run recipes for the OpenAI Agents output guardrail, a LangChain retriever, and a LlamaIndex reader. Every example runs against the shipped keyless + free-tier surfaces.

The installer ships with the demo key so you can try every client without signing up.