biomedmcp

io.github.unibio-intelligence由 GitHub 帳號 unibio-intelligence 發布v0.2.5更新於 Nov 6, 2026

Integrated public biomedical research with cited, bounded evidence.

已驗證Streamable HTTP可網頁執行知識與記憶資料與分析

概覽

AI 產生的概覽

BioMedMCP 讓 AI 助理以可查證引用的方式存取 PubMed、UniProt、ClinicalTrials.gov、ChEMBL、Open Targets、GTEx 等公開生物醫學資料來源。

可以做什麼
  • 依文字檢索基因、蛋白質、藥物、疾病、表型、通路、變異、研究、專利與作者,並依需求取得單筆記錄的指定證據章節
  • 跨 PubMed、Europe PMC、PubTator 與 LitSense 檢索論文,並依疾病、介入、期別與招募狀態檢索臨床試驗
  • 取得變異證據(ClinVar、gnomAD、CIViC 及選用預測來源),並分析已安裝的公開癌症基因體研究
  • 將名稱解析為帶物種的精確 ID,列出與某筆記錄關聯的記錄,並建立含引用的多章節檔案
  • 執行特定資料來源操作、平行批次執行最多十個讀取呼叫,並查看可用資料來源、章節、工作流程與剩餘配額
適用情境
適合需要以可查證、含引用的記錄回答生物醫學研究問題的情境,例如標的檔案、變異解讀、文獻回顧或臨床試驗查詢。
執行需求
可使用託管的 Streamable HTTP 端點(依提示登入),或本機安裝:Python 3.13+ 與 biomedmcp 套件(或容器映像),執行時設定 UBI_PROFILE 環境變數。部分資料來源需要金鑰或已安裝的資料集;憑證應放在環境變數或用戶端密鑰儲存中。
安裝前請注意
登入託管服務會將使用綁定至帳號,並受配額限制。部分資料來源需要 API 金鑰或已安裝資料集。取回的資料保留其來源條款;資料來源失敗或章節缺漏代表存在缺口,而非沒有證據。私有資料湖、專案匯入與建模不在此伺服器範圍內。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 biomedmcp,將其新增到工作區。
  2. 為需要使用其工具的對話啟用該服務。

Web executable,透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。

交給代理程式安裝

把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式,或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南,先向你展示發布者、端點和登入方式,經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。

閱讀 https://sourceweft.com/mcp/io.github.unibio-intelligence%2Fbiomedmcp/install.md,並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前,先告訴我它的發布者、端點和登入方式,等我確認。不要把 API 金鑰寫進命令或我會提交的檔案;問我金鑰應從哪裡取得。

檢視代理程式所依據的安裝指南

其他 MCP 客戶端

把它新增到你客戶端的 mcpServers 設定中。

{
  "mcpServers": {
    "biomedmcp": {
      "type": "http",
      "url": "https://biomedmcp.unibiointelligence.com/mcp"
    }
  }
}

README

biomedmcp

BioMedMCP gives your AI agent cited access to biomedical sources, including PubMed, UniProt, ClinicalTrials.gov, ChEMBL, Open Targets and GTEx, so its answers point to records you can check.

Start with a research question

Try this with your MCP client:

Build an evidence notebook on LDL-receptor biology and cholesterol lowering. Start with a human LDLR target dossier, inspect tissue expression, then find PCSK9 genetic studies. Separate observations, intervention results and open questions. Cite the records you read and identify any unavailable sources.

Connect to https://biomedmcp.unibiointelligence.com/mcp using Streamable HTTP and sign in when prompted. For a local server, install the Python distribution or build from this checkout; see connection options.

Run the same investigation locally

Use Python 3.13+ in an activated virtual environment:

sh
python -m pip install biomedmcp==0.2.5UBI_PROFILE=local biomedmcp build_dossier '{"workflow":"target_dossier","focus":"LDLR","species":9606}'

The single Python distribution includes all source adapters. From a source checkout, uv sync --frozen installs the locked environment; prefix commands with uv run --frozen when using it.

These focused calls can fill gaps in the notebook:

sh
UBI_PROFILE=local biomedmcp get_record '{"entity":"gene","id":"LDLR","species":9606,"sections":["expression"]}'UBI_PROFILE=local biomedmcp search_literature '{"query":"PCSK9","filters":{"source":"pubmed","date_from":"2024-01-01"}}'UBI_PROFILE=local biomedmcp get_variant_evidence '{"id":"PCSK9 R46L"}'

Check each response's source, species and identifiers before combining records; a partial or unavailable section is a gap, not a negative result. Follow its continuation and recovery links to see more than the preview.

Choose the next question

Research taskWhere to start
Separate target observations from causal claimsEvidence investigations
Inspect an allele without conflating transcript or genome buildTyped call examples
Read a paper's accessible text and tablesget_record with entity: "article" and sections: ["fulltext"]
Examine registered studiestrial_search; inspect each study's submitted record
Reproduce a local analysisSynthetic study walkthrough

Connect a local agent

sh
UBI_PROFILE=local biomedmcp-server --transport http --host 127.0.0.1 --port 8000

Use http://127.0.0.1:8000/mcp, or configure a stdio client using the connection guide. The call reference covers exact JSON arguments; result handling explains partial results and evidence boundaries.

Some sources need a key or installed dataset. Run biomedmcp diagnose and consult source access. Keep credentials in environment variables or client secret storage. A failed source doesn't mean there is no evidence.

The integration code uses Apache-2.0. Retrieved data retains its source terms. See notices, release notes and contributing.

Acknowledgements

We gratefully acknowledge support from Anthropic's AI for Science rare disease research grants.

來源:README.md,提交 70ca3db

工具

13

search_records

Search biomedical records

Search genes, proteins, drugs, diseases, phenotypes, pathways, variants, pharmacogenomics, GWAS, diagnostics, adverse events, studies, patents or authors by text; returns IDs for get_record (variants: get_variant_evidence). Papers: search_literature. Trials: search_trials. Patent search is chemistry-linked (SureChEMBL); assignee/inventor/CPC search is query_source(source=patents). e.g. {"entity":"gene","query":"BRCA1","filters":{"species":9606}}

輸入參數limitquery必填cursorentity必填outputfiltersfreshness

get_record

Get record and evidence sections

Fetch one record by ID with optional evidence sections, e.g. gene expression/eqtl/diseases/druggability, drug label/safety, article fulltext, trial eligibility, protein domains/structures. Variants: get_variant_evidence. List an entity's sections with get_capabilities(entity=...). e.g. {"entity":"gene","id":"PCSK9","species":9606}

輸入參數id必填entity必填sourcesspeciessectionsfreshnesssource_versionssection_parameters

search_literature

Search literature

Search papers across PubMed, Europe PMC, PubTator and LitSense with date, author, journal, article-type and open-access filters. Deduplicates across sources and reports each source's outcome. Read abstracts or full text with get_record(entity=article). e.g. {"query":"PCSK9 LDL cholesterol"}

輸入參數limitquery必填cursoroutputfiltersfreshness

search_trials

Search clinical trials

Search clinical trials (ClinicalTrials.gov, NCI) by condition, intervention, phase and recruitment status. Read arms, eligibility or outcomes with get_record(entity=trial). Does not assess an individual's eligibility. e.g. {"query":"familial hypercholesterolemia","filters":{"status":"RECRUITING"}}

輸入參數limitquery必填cursoroutputfiltersfreshness

get_variant_evidence

Get variant evidence

Get evidence for one variant (rsID, gene + protein change, HGVS or chr:pos:ref:alt). Defaults: ClinVar, gnomAD population, CIViC; add predictions (AlphaMissense), eve, cosmic, oncokb, hotspots, gwas or cbioportal. Ambiguous assembly or transcript is reported, not guessed; predictions keep source and uncertainty. e.g. {"id":"PCSK9 R46L"}

輸入參數id必填sourcesassemblysectionsfreshnesssection_parameters

analyze_study

Analyze genomics study

Analyze an installed public cancer genomics study: gene_profile, mutation_cohort, survival, compare, co_occurrence, filter_cohort, top_mutated. List installed studies with get_capabilities(source=study); ubi_synthetic is invented demo data. Report sample denominators and exclusions. Private data analysis uses the Ubi platform. e.g. {"analysis":"gene_profile","study":"ubi_synthetic","gene":"TP53"}

無需輸入參數。

resolve_entity

Resolve identifiers

Map a name, alias or ambiguous identifier to exact species-qualified IDs before joining evidence. Returns candidates when ambiguous; a provider match score alone is not identity. Skip when you already hold an exact ID. e.g. {"entity":"gene","query":"LDL receptor","species":9606}

輸入參數query必填entityspeciesassembly

build_dossier

Build research dossier

Build a multi-section dossier in one call: verified identity, default record sections and top related papers, with citations and per-source outcomes. Workflows: target_dossier, model_input_bundle (gene), drug_review, disease_landscape, pathway_review, variant_dossier, negative_evidence (bounded literature check; never proves absence). Add sections with get_record. e.g. {"workflow":"target_dossier","focus":"LDLR","species":9606}

輸入參數focus必填speciesassemblysectionsworkflow必填section_parameters

query_source

Query source directly

Run a source-specific operation that other tools do not cover, e.g. ChEMBL bioactivities, Ensembl coordinates, PDB structure quality, OmniPath interactions, NCBI Datasets genomes, ncbi-records for GenBank/GEO/SRA. Read get_capabilities(source=...) for operations and params first. e.g. {"source":"gtex","operation":"get_metadata","params":{"info_type":"tissues","dataset_id":"gtex_v10"}}

輸入參數paramssource必填freshnessoperation必填

run_batch

Run batched reads

Run up to ten independent read calls in parallel with one quota reservation; each result keeps its own outcome. build_dossier and utilities are not batchable. e.g. {"calls":[{"operation":"get_record","arguments":{"entity":"gene","id":"PCSK9","species":9606}}]}

輸入參數calls必填idempotency_key

get_capabilities

Inspect get_capabilities

List enabled sources, entity sections, dossier workflows and source operations with exact parameter schemas. Narrow with source or entity; diagnose=true probes source health. Examples: biomedmcp://guide. e.g. {}

輸入參數entitysourcediagnose

get_quota

Check quota

Show remaining research allowance and reset times; signed-in accounts get higher limits. e.g. {}

無需輸入參數。

版本歷史

2
  1. v0.2.5最新Nov 6, 2026
  2. v0.2.4Nov 5, 2026

評分

已滿足適用檢查項的 100%

該評分衡量此收錄條目的完整程度,依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價,也不表示伺服器安全或目前可用:可連通性是帶日期的檢查結果,評分也不包含安裝量或使用者評價。

  • 近 30 天內可連通必需25 分
  • 已宣告認證方式必需10 分
  • 已列出工具必需15 分
  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分