RBAi, Scholar, or MCP Cloud: choose by the work you need to do

Start with the task, not the acronym

The three names describe different entry points into research work. Choose by the next action you need to take, not by which product sounds most capable. This guide is based on the company’s product facts and public search and setup guides as of 4 October 2026. It is not a comparison benchmark or validation of generated answers.

The product overview gives the current presentation. This note deliberately does not duplicate prices, catalog totals, or daily allowances: check the product pages before designing around them.

RBAi: a research workspace, with alpha access by request

RBAi — Rhesus Bioagentic AI — is presented as a traceable agentic workspace for biological research questions. Its public status is alpha, and access is currently by request. The appropriate next step is Contact, not an assumption that the application is open to everyone.

When you request access, describe the biological question, the source material you need, and what traceability means for your work. For example: do you need original record identifiers, source links, or a way to inspect the basis of an answer? Check those requirements with the team before depending on the alpha for a workflow. This guide does not assert that any particular export, audit format, or experimental capability is available.

A research workspace is not evidence that a biological conclusion is correct. Any substantive claim still needs to be checked against the original source and the limits of the study.

RBAi Scholar: begin with literature discovery

Use RBAi Scholar when your next step is a reading list or a particular paper. The product describes an index of biomedical paper abstracts drawn from PubMed, bioRxiv, and medRxiv. Discovery over abstracts is not a substitute for reading the original paper, and a retrieved preprint should not be treated as a peer-reviewed result merely because it appears in search.

The query-syntax guide documents two explicit retrieval modes and a quoted form of semantic search:

• Semantic: write a sentence describing the question when you want meaning-based discovery and do not yet know the literature’s vocabulary. A close match is a candidate to inspect, not a confidence score for the claim.
• Keyword: use known terms or identifiers. This is the mode in which AND, OR, NOT, and -term are interpreted as operators. Bare terms are combined with AND.
• Semantic with quoted requirements: put a required string in quotation marks inside the question. The guide calls this hybrid search; it is reached through a semantic query, not a separate mode parameter.

For example, the following is a documented form of a keyword request, not a report of returned papers:

{
  "mode": "keyword",
  "query": "TP53 AND apoptosis NOT mouse"
}

The exclusion applies to indexed words. It is not a validated filter for study species: a record may omit a species from its abstract or mention mouse work in a different context. Inspect the study before drawing that conclusion.

For semantic discovery with an explicit requirement, the guide uses:

role of "MDM2" in stabilising mutant p53

Do not write "TP53" OR "BRCA1" in semantic mode expecting Boolean OR. The documented behavior treats the two quoted strings as requirements together, while OR is ordinary text. Use keyword mode for that Boolean intent. Parenthesized grouping is not supported by the current guide.

Scholar also offers REST API and MCP access. Choosing it is about the literature-search task, not a requirement to stay in the web interface.

MCP Cloud: connect a client to biological-data tools

Use MCP Cloud when you are building or operating a workflow that needs biological-database tools from an MCP client or agent. Its setup page describes a managed endpoint and a catalog that you can inspect before connecting a client.

The published interface uses two tool definitions: find_tools discovers an appropriate catalog tool from a plain-language description, and call_tool invokes a selected tool. Read the selected tool’s input requirements and preserve the underlying database record identifiers. Finding a tool does not establish that its output answers your biological question.

Start with the catalog, then use the API-key and client setup for the configuration actually supported by your client. Keep the key out of public source code, browser examples, screenshots, and shared logs. A public identifier is a safer starting point for a setup check than confidential research material or identifiable patient information.

Do not assume that MCP access removes a database’s coverage, versioning, or interpretation limits. Record the database, tool, arguments, retrieval date, and source version where available. If a version is not exposed, say that it was unavailable rather than inventing one.

A workflow that keeps the evidence attached

1. Write the information need. Specify the organism, entity, mechanism, or study type you care about before choosing a mode or tool.
2. Discover the literature. Use Scholar when the missing piece is a paper or a reading list; save the query, mode, and source identifiers.
3. Retrieve the needed biological records. Use MCP Cloud when the next step needs developer-facing access to a biological database. Check the tool description rather than relying on its name alone.
4. Discuss workspace access if needed. Request RBAi access if you want to explore the agentic workspace. Confirm the alpha’s fit for your requirements rather than assuming an automatic hand-off between products.
5. Check the originals. Separate relevance, source quality, and support for a specific claim. A useful search hit or successful tool call is not experimental validation.

This is a suggested decision path, not an executed end-to-end experiment. If you want to evaluate literature discovery systematically, use the biomedical-search evaluation checklist to prepare the query set and judgement process first.

Check current access before you build

Product access policies and limits can change independently of this article. Use RBAi Scholar and MCP Cloud for their current setup and plan details; use Contact for RBAi access or help matching a research workflow to an entry point.

For search behavior, keep the query-syntax guide beside your implementation. It is more actionable than assuming that every mode interprets the same words in the same way.