FreeHow credits work →

Academic MCP for Antigravity

Add ScholarMuse to mcp_config.json and click Authenticate. Agents in Antigravity can then use official statistics, compound and trial records and archive material in the work they carry out.

How to connect ScholarMuse to Antigravity

ScholarMuse is added as a custom server in Antigravity’s MCP configuration file. You can also paste “Read https://scholarmuse.ai/install/agent.md and follow the instructions to connect ScholarMuse” to an agent and let it edit the file.

Full connect guide →
  1. Step 1

    Open the MCP configuration

    In the Antigravity IDE, open the “…” menu at the top of the agent panel, choose MCP Servers, then Manage MCP Servers and View raw config. This opens mcp_config.json. You can also open the file directly at ~/.gemini/config/mcp_config.json.

  2. Step 2

    Add the server entry

    Add this to the file and save. The field must be named serverUrl; Antigravity does not accept url or httpUrl for remote servers. If the file already lists other servers, add only the scholarmuse entry under mcpServers.

    {
      "mcpServers": {
        "scholarmuse": {
          "serverUrl": "https://scholarmuse.ai/mcp"
        }
      }
    }
  3. Step 3

    Click Authenticate

    Open Settings, go to the Customizations tab and click Authenticate next to scholarmuse under Installed MCP Servers. Antigravity registers itself with the server automatically, so there is nothing else to fill in.

  4. Step 4

    Approve in your browser

    A scholarmuse.ai window opens. Sign in, approve access and choose which discipline packs Antigravity may use. If the page then shows an authorization code, copy it into the Antigravity settings panel and click Submit. There is no API key to create or paste.

  5. Step 5

    Check that it works

    Ask an agent to list the ScholarMuse capabilities. By default Antigravity asks for your approval before it runs an MCP tool you have not configured.

What Antigravity agents can use once connected

Antigravity hands a whole task to an agent, so this list mixes inputs the task needs, from economics, chemistry, medicine and history, with a figure for the report at the end.

Economic & social statistics
An agent building a panel dataset fetches each official series itself and records where it came from.
Compounds & materials
Compound identifiers and properties, or crystal structures and band gaps, as the starting table for a screening task.
Drugs & targets
Targets, bioactivity and adverse events from ChEMBL and openFDA, for a pharmacology write-up that cites its records.
Clinical trials search
A structured list of registered trials by condition, intervention, phase and status, ready to be tabulated.
Historical archives
Digitised newspapers, manuscripts and images for a text-analysis or digital humanities project.
Find datasets
The agent can find a published dataset and report its licence before it plans the rest of the task around it.
Search papers
Background reading for the task plan, taken from indexed papers with DOIs.
Scientific figure maker
When the analysis is done, the same agent can produce a schematic of the study design for the report.

Things to ask Antigravity

Economic & social statistics

You say to your agent

Use ScholarMuse to build a panel of annual GDP per capita and unemployment for the 27 EU countries from 2005 to 2024. Save it as panel.csv and write a short note listing the source of each series.

Series with unit, source and licence for each country, which the agent assembles into one file.

Compounds & materials

You say to your agent

Look up the known polymorphs of ZnO and SnO2 with their band gaps and formation energies, put them in a table, and plot band gap against formation energy.

Material records with structure, band gap and stability values and the database each one came from.

Clinical trials search

You say to your agent

List phase 2 and 3 trials of psilocybin for depression registered since 2019, with status, enrolment and sponsor, and summarise the designs in a table.

ClinicalTrials.gov entries with registration number, phase, status and sponsor.

Frequently asked questions

Most editors name the address field url. Antigravity’s documentation requires serverUrl for remote servers and says url and httpUrl are not supported. Rename the field, save the file and authenticate again.

Do not rely on finding it there. Add it as a custom server through mcp_config.json as shown above; once saved, it appears under Installed MCP Servers with the other servers.

Yes. Antigravity reads a global file at ~/.gemini/config/mcp_config.json and a workspace file at .agents/mcp_config.json. Put the same entry in the workspace file to limit it to that project.

No. Those fields are for servers without dynamic client registration. ScholarMuse supports it, so the entry holds only the address, and you sign in at scholarmuse.ai. A new account gets 50 credits once.

Every call your agent makes uses credits, and a call is charged only when it succeeds. There are no subscriptions. A new account starts with a one-time gift of 50 credits. The 6 web tools under /tools are free without an account, up to 5 runs a day.

Requests are used only to answer you. We do not train models on them.

Ready to make your agent a better scholar?

Connect once. Papers, data and figures from 50+ sources, in the agent you already use.

Works with compatible AI agents

ClaudeChatGPTCodexCursorWorkBuddyTraeGemini CLIAntigravityQoderCodeBuddyKimi CodeQwen CodeClineKiroWindsurfZedOpenCodeOpenClawDifyCherry StudioLM StudioGooseAmpContinueRoo CodeVS CodeJetBrainsWarpRaycastCopilotPerplexityAny MCP client

Data from OpenAlex · Crossref · PubMed · arXiv · NIH · UniProt · PDB · PubChem · NASA · World Bank · USGS · GBIF and more