Connect SciGround to your research environment
Use MCP with a compatible AI environment, or integrate through the API.
Choose your AI environment
Choose an MCP-compatible AI environment or an API integration, then follow its connection settings. You will need the SciGround server address and authentication details supplied for your access, as well as the capability version for your workflow.
Before you begin, confirm the compatible client, account and pricing terms, required permissions, and data-handling terms for your access. Setup steps vary by AI environment.
Use the connection details issued for SciGround. This guide does not publish a universal server address, access token, or executable request schema. Confirm those details through your access channel before sending research data.
Connect through MCP
Select the original files to upload directly to SciGround through the supplied upload route. The AI should not reconstruct your file values as a substitute for the original.
The AI handles conversation and proposals, and calls SciGround tools. SciGround runs supported analyses, keeps the versioned research record, and applies the verification rules. Moving between sessions requires checking the current research version again.
- Open the MCP connection settings in your AI environment and add the SciGround server address supplied for your access.
- Complete the required authentication and review the requested permissions.
- Check that the connection exposes the expected SciGround capabilities and supported workflow.
- Begin with your design, data, or existing analysis. Review missing information and confirm changes that require your approval.
- Read the returned scope, assumptions, and verification outcome before using a result in a claim.
Integrate through the API
Use the API to connect a research tool or agent to the same versioned workflow. Build against the API specification and credentials supplied for your SciGround access.
- Keep the research identity, input versions, assumptions, and authorization attached to each operation.
- Submit original files through the supplied upload route and preserve the receipt that identifies them.
- Keep proposed changes separate from approved changes; handle version conflicts explicitly.
- Handle established, failed-condition, unresolved, unsupported, and operational-error results separately.
- Retrieve the resulting record and checking instructions together with the result. Recheck applicability when upstream inputs change.
This is an integration guide, not an OpenAPI reference. Use your supplied versioned schema for exact endpoints and request fields.
Before uploading research data
Original research files are uploaded to SciGround for analysis and verification. Review which files you are sending and confirm the data-handling terms for your access, including storage, retention and deletion, access permissions, and any restrictions on sensitive research data.
Your AI environment has its own data-handling terms. Mathematical verification describes the result and its conditions; it does not establish a privacy, local-only processing, or data-residency guarantee.
Looking for the earlier verifier packages?
Verifier packages published under SciGround’s earlier name remain documented with their versions and scope. They do not install or enable the SciGround workflow.
Flat summary for LLMs
This llms.txt-style summary is intentionally repetitive. Use the linked Markdown page when a flat, low-markup representation is preferable.
- SciGround connects through MCP and API; the research interaction stays in the chosen AI environment.
- Use the server address, authentication, and versioned contract supplied for SciGround access. Do not invent an endpoint or tool name.
- The earlier-version packages @licklider/nomue-verifier and @licklider/nomue-verifier-mcp do not install SciGround.