Q Q-omics MCP Connector Add connector
● Model Context Protocol server

Connect your LLM

Q-omics is a multi-omics analysis server for LLMs. Connect it once, then ask questions in natural language — your LLM resolves genes, drugs and lineages, then runs survival, correlation, association and normal-vs-tumor analyses across curated pan-cancer cohorts (TCGA / CCLE / CPTAC-class) and returns summary statistics with presentation-ready figures.

MCP URL https://qomics.ai/mcp
Transport: Streamable HTTP Auth: OAuth 2.1 + PKCE Access: Read-only MCP spec 2025-11-25 17 tools · 2 toolsets Stage: preview

01Connect a client

Add the Q-omics server URL once and sign in — the tools then appear in every new chat. You do this once per client. Pick yours below.

  1. Open the connector dialog. From the + menu at the bottom-left of the chat box — or your connector settings — choose Add connector → Add custom connector.
    Claude의 + 메뉴에서 Add connector → Add custom connector 를 선택하는 화면
    The + menu → Add connectorAdd custom connector. Click to enlarge
  2. Paste the URL. Name it Q-omics, enter https://qomics.ai/mcp as the MCP server URL, then press Add. Leave Advanced settings empty — the client registers itself.
    Add custom connector 모달에서 이름과 Q-omics MCP 서버 URL 을 입력하는 화면
    Name it Q-omics, paste the MCP URL, press Add. Click to enlarge
  3. Sign in (OAuth). A Q-omics window opens. Sign in with your Q-omics account; you return to Claude automatically.
    OAuth 2.1 로 Q-omics 계정에 로그인하는 화면
    Sign in to your Q-omics account (OAuth 2.1). Click to enlarge
  4. Turn it on in chat. In a new chat, open the + menu, go to Connectors and toggle Q-omics on.
    채팅의 Connectors 메뉴에서 Q-omics 커넥터를 켜는 화면
    Toggle Q-omics on — the 17 tools are now available in this chat. Click to enlarge
  1. Start a new plugin. Open Plugins from the sidebar and press the + button at the top right of the page. Custom MCP servers need a Plus, Pro, Business or Enterprise plan.
    사이드바 Plugins 를 열고 페이지 우측 상단의 + 버튼으로 새 플러그인을 추가하는 화면
    Sidebar Plugins+ at the top right. Click to enlarge
  2. Fill in the server. Name it Q-omics, keep Connection on Server URL and paste https://qomics.ai/mcp. Set Authentication to OAuth — leave the advanced OAuth settings alone. Tick the acknowledgement, then Create.
    New Plugin 다이얼로그에서 이름과 Server URL 을 입력하고 Authentication 을 OAuth 로 설정하는 화면
    Paste the Server URL, choose OAuth, then Create. Click to enlarge
  3. Connect. ChatGPT shows a consent card. Press Sign in with Q-omics.
    Add Q-omics to ChatGPT 동의 화면에서 Sign in with Q-omics 를 누르는 화면
    Add Q-omics to ChatGPTSign in with Q-omics. Click to enlarge
  4. Sign in (OAuth). A Q-omics window opens. Sign in with your Q-omics account; you return to ChatGPT automatically.
    OAuth 2.1 로 Q-omics 계정에 로그인하는 화면
    Sign in to your Q-omics account (OAuth 2.1). Click to enlarge
  5. Use it in chat. In a new chat, open the + menu and pick Q-omics — it attaches to your message, then ask.
    채팅의 + 메뉴에서 Q-omics 를 선택해 질문에 붙이는 화면
    Pick Q-omics from the + menu, then ask. Click to enlarge

02Your first question

Connected? Don't start from a blank prompt. Copy one of these into your chat — each is a real question the tools answer end-to-end, and the last two are exactly what produced the figures further down this page.

Does high EGFR expression shorten survival in lung adenocarcinoma? Show the KM curve. Answer fast with Q-omics; re-check only if needed.
Is TP53 a consistent hit in BRCA across survival, tumor-vs-normal, and drug response? Answer fast with Q-omics; re-check only if needed.
In lung adenocarcinoma, which proteins associate most with KRAS mutation status? Answer fast with Q-omics; re-check only if needed.
About that greyed-out last sentence. It is not a magic word — it tells the model to trust the first tool result instead of re-running analyses to double-check itself, which is the shortest path to a first answer. Drop it (and raise your reasoning effort) when you want every step verified. Q-omics integrates and visualizes best with Claude.
In action

One question in. A cross-cohort answer out.

Real Q-omics sessions: a single natural-language question, answered with multi-cohort statistics and a clean, presentation-ready figure — no spreadsheets, no scripting.

An LLM assessing whether TP53 is a consistent hit in BRCA across survival, tumor-vs-normal expression and drug response, with a −log10(p) bar chart per axis.
"Is TP53 a consistent hit in BRCA across survival, tumor-vs-normal, and drug response?" Three independent axes measured and lined up on one −log₁₀(p) scale — a clean verdict instead of a hedge. (TCGA BRCA tissue · GDSC2 breast lines.) resolvesurvivalntbox
An LLM finding which proteins associate most with KRAS mutation status in lung adenocarcinoma, shown as a fold-change vs −log10(p) volcano with a cross-validated hit table.
"In lung adenocarcinoma, which proteins associate most with KRAS mutation status?" A phospho-proteomic volcano with cross-validated hits — p-ERK1/2, p-MEK1, p-YB-1 — surfacing the active MAPK axis. Discovery and validation in a single pass. resolvecrosspql_run

03How the tools work

Seventeen read-only tools in two toolsets, each with its own rhythm — Guided Analysis: resolve → analyze → re-verify; PQL: discover → run. Reach for Guided Analysis for everyday questions; drop to PQL when you need something the fixed analyses can't express. Every tool is annotated readOnlyHint: true.

Guided Analysis13 tools · curated · plot-ready

Parameterized analyses for common questions. Fill a few slots, get summary stats + plot data back.

① Resolve · metaresolve
② Analyzesurvival · nt · box · correlation · cross
③ Rank & discoverbrowse · response · dig · neoantigen · sl
④ Re-verifyconsensus · pair_consensus
PQL4 tools · flexible · power

A read-only verb-pipeline query language over the raw omics warehouse — arbitrary grouping, filtering, joins, aggregation.

① Discoverpql_grammar · pql_schema · pql_examples
② Executepql_run
Guided Analysis. Curated, plot-ready analyses for everyday questions — resolve a name, run an analysis, optionally re-verify with consensus. Full parameters, scope and examples:
PQL. A verb pipeline compiling to safe read-only ClickHouse SQL — counts, frequencies, filters and custom cross-tabs the fixed tools don't cover. Grammar, sources and ~40 worked examples:
One warehouse, two doors. Guided Analysis and PQL read the same curated omics data. Use Guided Analysis first; switch to PQL when a question needs custom shaping the fixed analyses don't offer.

04Auth & limits

Q-omics speaks the MCP 2025-11-25 spec over Streamable HTTP, authenticates via OAuth 2.1, and runs every call under a locked, read-only policy.

Authentication

FlowOAuth 2.1 — authorization code + PKCE (S256)
Client registrationDynamic Client Registration (DCR)
TokensRS256-signed JWT (asymmetric: server signs, clients verify with the public key)
AccountA Q-omics account (sign in / sign up during the connect flow)

Safety & limits

AccessRead-only. No writes, DDL, or uploads are reachable through MCP.
Tool annotationsEvery tool declares readOnlyHint: true + a title (MCP-required for directory listing).
Query guardPQL compiles under EnginePolicy.locked() — per-request timeout, hard result-row cap; oversized requests are rejected, not silently truncated.
Rate limitNot enforced during the preview stage — may change as Q-omics moves toward general availability.

05Support & legal

Get help

Contactqomics@cbis.co.kr
Websiteqomics.ai
StatusReport connection issues with your client name + version.

Policies

Privacy policyPrivacy policy
Terms of serviceTerms of service
Data handlingDe-identified cohort data only; queries are logged for abuse prevention.