GLG

It is https://mcp.glgresearch.com/myglg/mcp as of 2026-08-23. Anthropic's directory previously published a different host, access.glgresearch.com/mcp-myglg-gateway/mcp, and that older endpoint still answers requests. Only the new one is listed now, so a manually configured client pointing at the old host should be repointed.

Yes. GLG states that the MCP connector requires an active GLG account, and directs anyone without one to their GLG account manager. There is no self-serve signup path and no trial tier documented for the connector. GLG sells to institutions, so the account you sign in with is the myGLG account your firm already holds.

Because connector access is enabled per account by GLG at the account owner's request. GLG's troubleshooting section explains that sign-in and connection succeed before enablement, and the instruction only appears on your first data query. Once GLG turns it on, queries work immediately and you do not need to reconnect.

Two GLG role flags, not capability grants. The server's descriptor advertises urn:glg:session-access-flags/role-glg:2 and urn:glg:session-access-flags/role-myglgmcp:1. Each names a role and a level rather than an object or a verb, so neither one separates reading a transcript from registering you for an event.

No tool in the published list does that. The 13 names cover reading your account plus one event RSVP. GLG markets a future Match step that would let you request a call in conversation, but describes it in the future tense, and no tool named in Anthropic's directory corresponds to it.

No. Unlike Third Bridge, which publishes standalone Pilot MCP Terms, GLG publishes no MCP-specific client agreement. Its existing Terms of Use, Usage Policies and GLG Library Terms govern the connector, and GLG states your Client Agreement controls wherever it conflicts with the standard terms.

Partly. GLG states the data returned to answer a question is fetched live and discarded, but that it logs the tool requests your assistant makes, including parameters like search terms and call IDs, for product analytics. GLG states those logs do not capture your prompts or your assistant's responses, and are not used to train any model.

No. GLG's documentation states the connector cannot access your conversation history, the prompts you type, or the responses your assistant generates. It also cannot reach another GLG account's data or your billing information. GLG separately states that your GLG data is never used to train or improve any AI model.